Two novel predictive biomarkers for osteosarcoma and glycolysis pathways: A profiling study on HS2ST1 and SDC3

Introduction:

Prognostic biomarkers for osteosarcoma (OS) are still very few, and this study aims to examine 2 novel prognostic biomarkers for OS through combined bioinformatics and experimental approach.

Materials and methods:

Expression profile data of OS and paraneoplastic tissues were downloaded from several online databases, and prognostic genes were screened by differential expression analysis, Univariate Cox analysis, least absolute shrinkage and selection operator regression analysis, and multivariate Cox regression analysis to construct prognostic models. The accuracy of the model was validated using principal component analysis, constructing calibration plots, and column line plots. We also analyzed the relationship between genes and drug sensitivity. Gene expression profiles were analyzed by immunocytotyping. Also, protein expressions of the constructed biomarkers in OS and paraneoplastic tissues were verified by immunohistochemistry.

Results:

Heparan sulfate 2-O-sulfotransferase 1 (HS2ST1) and Syndecan 3 (SDC3, met all our requirements after screening. The constructed prognostic model indicated that patients in the high-risk group had a much lower patient survival rate than in the low-risk group. Moreover, these genes were closely related to immune cells (P < .05). Drug sensitivity analysis showed that the 2 genes modeled were strongly correlated with multiple drugs. Immunohistochemical analysis showed significantly higher protein expression of both genes in OS than in paraneoplastic tissues.

Conclusions:

HS2ST1 and SDC3 are significantly dysregulated in OS, and the prognostic models constructed based on these 2 genes have much lower survival rates in the high-risk group than in the low-risk group. HS2ST1 and SDC3 can be used as glycolytic and immune-related prognostic biomarkers in OS.

1. Introduction

Osteosarcoma (OS), a highly malignant primary tumor of bone, shows a generally poor prognosis for patients who develop a recurrent disease or have primary metastases.^[1] It is most common in children and young adults and usually metastasizes easily to the lungs. Its etiology is multifaceted, and its specific pathogenesis remains poorly understood.^[2] Treatment of OS is primarily based on surgical interventions and chemotherapies targeting the focal area; however, several unresolved issues such as the poor prognosis of patients with recurrent OS and non-removal of some OS using surgery^[3] remain. Although surgery combined with chemotherapy can greatly improve the prognosis of patients with OS, the prognosis of metastatic or recurrent OS remains unsatisfactory, and immunotherapy has proven to be a very promising modality for the treatment of human malignancies.^[4] In recent years, with the advancement of bioinformatics, more and more studies on biomarkers of tumors have been reported by scholars.^[5] As a highly malignant tumor, there is an urgent need to identify biomarkers that are closely related to prognosis in order to guide clinical diagnosis, treatment, and prediction of prognosis.

Immunotherapy is being recognized as a novel tool to treat tumors and accepted by an increasing number of scholars.^[6–8] Several signaling pathways play a crucial role in the homeostasis of the highly specific internal immune environment o miR-29c f a bone.^[9] Expression of the transcription factor Forkhead Box P3 regulatory T (Treg) cells has been shown to play a key role in maintaining immune homeostasis and preventing autoimmunity in vivo.^[10] It has been shown that chimeric antigen receptor T-cell therapy can revolutionize the clinical treatment of hematologic malignancies and that it can lead to significant advances in the treatment of the hematologic system.^[11] In addition, chimeric antigen receptor (CAR)-T cells are capable of producing immune responses to a variety of cancers, including non-small cell lung cancer, and it shows great promise as a novel agent for immunotherapy, CAR-T cells in non-small cell lung cancer for the treatment of cancer.^[12] More interestingly, some researchers found that in lung cancer, the proportion of resting NK cells, monocytes, M0 macrophages, resting mast cells, eosinophils and neutrophils was significantly lower in tumor tissue than in normal tissue.^[13] Interleukin (IL)-9-producing CD4+ T cells in humans are thought to represent a distinct subpopulation of T helper cells called IL-9, and in recent years, there has been increasing evidence that IL-9-producing T cells may have therapeutic utility in the eradication of advanced tumors, particularly melanoma.^[14] Immunotherapy of cancer with CAR-T cells has significant clinical potential for malignancies, and successful regimens and strategies for CAR T-cell therapy can improve the efficacy and safety of such cancers.^[15] Numerous studies have shown that immune cell imbalance plays an integral role in the study of cancer. Immune cells have been shown to be dysregulated in a variety of tumors and it may be an integral factor in cancer progression.^[16,17] Hence, in order to gain more insight into the immune cell typing and deeply dissect the composition of immune cells in OS, we performed an analysis using the CIBERSORT^[18] software.

Glycolysis: Glucose participates in this pathway in an anaerobic or hypoxic environment, when carbohydrates are broken down into lactic acid, and a small quantity of adenosine triphosphate (ATP) is produced.^[19] It has also been found that the glycolytic enzyme lactate dehydrogenase A is capable of signaling via phosphatidylinositol 3-kinase in CD8+ T effector cells and that, in turn, removal of lactate dehydrogenase A inhibits phosphatidylinositol 3-kinase-dependent phosphorylation of Akt and its transcription factor target Foxo1, leading to antimicrobial immunodeficiency.^[20] Altered energy metabolism, one of the hallmarks of cancer, is a key feature that preferentially uses glycolysis to produce ATP or energy. Together, the oncogenic regulation of glycolysis and the multifaceted role of glycolytic components highlight the biological significance of tumor glycolysis.^[21] Li et al^[22] found a strong association between enhanced glycolytic activity and promotion of tumor immunity in breast cancer. Interestingly, studies have shown that the glycolysis-related genes GLCE and TPI1 can serve as biomarkers of prognosis in Ewing’s sarcoma and exhibit a dysregulated state of naive B cells, CD8+ T cells, activated NK cells and M0 macrophages in Ewing’s sarcoma.^[23] It was also found that cell subsets producing IL-17 or γδ T interferon-γ (IFN-γ) have fundamentally different metabolic requirements, with IFN-γ + γδ T cells being almost exclusively dependent on glycolysis and IL-17 + γδ T cells being able to participate in oxidative metabolism.^[24] In addition, it has been shown that the level of UBR5 is strongly correlated with the malignant phenotype and shorter survival of pancreatic cancer patients, and functional analysis has shown that UBR5 promotes the growth of pancreatic cancer cells by inducing aerobic glycolysis.^[25]

It is because of the poor prognosis of OS and the current paucity of biomarkers that can effectively predict prognosis. Therefore, we targeted an in-depth analysis of glycolysis and immune-related prognostic biomarkers based on multiple precise bioinformatics approaches to predict the prognosis of OS patients. Finally, we validated protein expression in OS and paraneoplastic tissues by immunohistochemical analysis.

2. Materials and Methods

2.1. Data download and cleaning

Gene expression profiles and the corresponding clinical data for OS were downloaded from the UCSC Xena database (http://xena.ucsc.edu/). Normal control samples were downloaded from the GTEx (https://www.gtexportal.org/home/) database. The glycolysis-related gene sets were downloaded from the Gene Set Enrichment Analysis database, from which the human glycolysis gene sets were selected. Statistical analysis, data cleaning, and graph plotting were performed by the R programming language, version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria; https://www.r-project.org/), and all expression spectrum data were log2 transformed. We normalized the data downloaded from 2 different databases to enable them to be analyzed at the same level. We excluded samples with incomplete clinical information and those not diagnosed with OS.

2.2. Differentially expressed genes (DEGs) analysis, gene ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis

To better understand the function of differentially expressed genes in OS and paraneoplastic tissues, we performed differential gene analysis and visualized the results using the “pheatmap” package and “ggplot2” package for gene expression profiles, with cut off values set to |logFC| > 0.5, fast discovery rate (FDR) < 0.05. We used the “org.hs.eg.db” package, Disease Ontology Semantic and Enrichment (DOSE) package, “clusterProfiler” and “enrichplot” packages to perform and visualize GO and KEGG pathway enrichment analysis on DEGs.

2.3. Construction of a prognostic model and risk assessment for OS

We took the intersection of upregulated DEGs with glycolysis-related genes to obtain corresponding DEGs. We first performed univariate Cox regression analysis of the genes of the intersection obtained in the previous step with survival status and survival time. P value < .05 was considered statistically significant. Second, to further refine the model and obtain the optimal value, a penalty function using the least absolute shrinkage and selection operator regression method was constructed. Finally, we performed multivariate Cox regression analysis to refine the model further, and P value < .05 was considered statistically significant. Based on the OS prognostic model, risk-scores of all patients were ranked in descending order, with those with risk scores below the mean considered to be the low-risk group and those with risk scores greater than or equal to the mean considered to be the high-risk group. We constructed risk score, survival, and heat maps according to the level of the risk score.

2.4. Survival analysis

We analyzed the survival information of all OS patients from 2 different perspectives. First, all patients were stratified into gene high expression group and gene low expression group based on the high and low expression of heparan sulfate 2-O-sulfotransferase 1 (HS2ST1) and Syndecan 3 (SDC3) genes, and the differences between the 2 groups were compared. Subsequently, we obtained a risk score for each patient based on the constructed prognostic model, and we classified cases above the median risk score as high-risk group and cases less than or equal to the risk score as low-risk group. We analyzed the survival differences of the prognostic model based on the division of high and low-risk groups.

2.5. Calibration of OS prognostic model

We constructed prognostic models for 1, 2, and 3 years’ receiver operating characteristic curve (ROC) diagnostic curves using the “survival,” the “survminer,” and the “timeROC” packages. The expression of these 2 genes was analyzed in relation to the high- and low-risk groups using the “ggpubr” package based on the 2 genes constructed for the model. We analyzed the principal components of the high and low-risk groups using the “ggplot2” package. Finally, nomogram-predicted calibration plots were drawn using the “rms” package to evaluate the difference between their predicted and actual values; column line plots were constructed based on the prognostic model to predict the OS patient survival probability at 1, 2, and 3 years

2.6. Immune cell typing analysis

To examine the immune cell composition of OS, gene expression profiles were analyzed using CIBERSORT software. All gene expression data were compared with the reference values set by the CIBERSORT software to derive the immune cell composition in 22 immune cells for each sample, and a P value < .05 was considered statistically significant. Then, we analyzed the immune cell composition of HS2ST1 and SDC3 genes in OS based on the constructed model and also studied their expression trends in relation to immune cells.

2.7. Correlation analysis of modeled genes and drug sensitivity

We have carried out further analysis at the level of drug sensitivity in order to analyze at multiple levels the genes that construct prognostic models for OS. We downloaded information on gene and drug sensitivity from the CellMiner^TM database (version: 2021.1, database:2.6, https://discover.nci.nih.gov/cellminer/home.do). After collating the data, we used the “impute” package, the “limma” package, the “ggplot2” package and the “ggpubr” package to carry out a detailed analysis of the 2 genes and drug sensitivities for which the models were constructed.

2.8. Immunohistochemistry

To confirm the protein expression of HS2ST1 and SDC3 in OS and paraneoplastic tissues, the expression of these 2 genes was analyzed using immunohistochemical staining. The pathological sections for immunohistochemistry were excised specimens obtained for pathological testing during surgery at the First Clinical Affiliated Hospital of Guangxi Medical University. The study followed the Declaration of Helsinki and was approved by the ethics department of the First Clinical Affiliated Hospital of Guangxi Medical University, and immunohistochemical staining analysis was performed on anonymous tissue specimens. Informed consent was waived by Department of Ethics, the First Clinical Affiliated Hospital of Guangxi Medical University. We used Proteintech (https://www.ptgcn.com/, HS2ST1, item no. 12453–1-AP; https://www.ptglab.com/SDC3, item no. 10886–1-AP) for labeling of specific protein expression regions. Fresh xylene was first used to soak the pathological tissue sections, followed by soaking in ethanol, washing the sections with sterilized pure water, followed by soaking of the sections in 10% formalin for 10 minutes. Then microwave antigen repair, blocking, primary antibody incubation, secondary antibody incubation, and fluorescent staining were performed to amplify the signal and finally for sealing of the sections. The stained sections were placed under an inverted microscope to evaluate protein expression in them. We performed immunohistochemical staining of 4 pairs (OS and paraneoplastic tissue) of pathological tissue sections for each gene, for a total of 16 sections. Subsequently, we used Image J software to calculate the positive regions for specific staining, and we imported the results of the calculations into IBM SPSS Statistics 25. We performed statistical analysis of the positive rate of OS tissue and the positive rate of paraneoplastic tissue using a t test of 2 paired sample means. Finally, we use GraphPad Prism 8 to visualize the statistical results.

3. Results

3.1. Basic information

After screening, eighty-eight samples with complete clinical data on OS were downloaded from the UCSC Xena database. A total of 396 skeletal muscle/muscle tissue samples, mostly normal controls, were downloaded from the GTEx database. A total of 317 human-associated glycolytic genes were downloaded from the Gene Set Enrichment Analysis database.

3.2. Analysis of DEGs, GO, and KEGG pathways

There were 54,751 genes in the expression profile of OS. Of them, 3562 genes remained after a DEG analysis, which were visualized as a heat map (Fig. 1A) and a volcano map (Fig. 1B) and analyzed in a more in-depth manner. We used the R programming language to learn more about their GO functions. GO Terms (Fig. 2A) are mainly enriched in glycosaminoglycan metabolic process, aminoglycan metabolic process, aminoglycan biosynthetic process, glycosaminoglycan biosynthetic process, etc. Figure 2B shows the KEGG pathway mainly enriched in the biosynthesis of glycosaminoglycans-heparan sulfate/heparin, glycolysis/gluconeogenesis, and insulin resistance.

Figure 1.

Heat map and volcano plot of differentially expressed genes. (A) A heat map of the top 100 differentially expressed genes, with red squares indicating up-regulated genes and green indicating down-regulated genes. (B) The red dots in the volcano plot indicate the up-regulated genes and the green dots indicate the down-regulated genes.

Figure 2.

GO enrichment analysis and KEGG enrichment analysis of glycolysis-related differentially expressed genes. (A) The left side of the ellipse shows the logFC value, and each color on the right side indicates a GO entry. (B) The red color of the innermost circle indicates the logFC value of each gene, and each color of the outer circle indicates a KEGG pathway. GO = gene ontology, KEGG = Kyoto Encyclopedia of Genes and Genomes.

3.3. OS prognostic model and risk assessment

To further understand the association between glycolysis-related genes and the prognosis of OS, 47 glycolysis-related DEGs after taking the intersection of upregulated DEGs and glycolysis-related genes were obtained (Fig. 3D). Subsequently, the genes were filtered and analyzed using univariate Cox regression analysis (Table 1), and only 5 genes fulfilled our screening criteria. The prognostic model was further finetuned using a more advanced least absolute shrinkage and selection operator regression analysis, and only 4 genes met our criteria (Fig. 3A and B). Finally, a more comprehensive multivariate Cox regression analysis (Fig. 3C and Table 2) was performed that revealed that only 2 remaining genes (HS2ST1 and SDC3) met our requirements. We developed a prognostic model for OS based on these 2 genes, determined individual risk scores for each patient, and categorized the patients into a high-risk group and a low-risk group. We constructed risk maps (Fig. 4A), survival maps (Fig. 4B), and heat maps of gene expression (Fig. 4C) for OS patients.

Figure 3.

LASSO regression, multi-factor COX regression, and Veen plots. The minimum penalty coefficients of the constructed LASSO regression model are shown in (A) and (B). The results of multi-factor COX regression are shown in (C). The intersection of differentially expressed genes with those of glycolysis is shown in (D). LASSO = least absolute shrinkage and selection operator.

Table 1.

Univariate regression analysis. Specifics of the univariate COX regression analysis.

ID	HR	HR.95L	HR.95H	P value
CDK1	1.752635	1.083551	2.834872	.022194
EXT2	0.456153	0.232635	0.89443	.02233
HS2ST1	0.438778	0.218688	0.880368	.020417
SDC3	0.623488	0.449581	0.864667	.004633
G6PD	0.432279	0.256107	0.729637	.001689

HS2ST1 = heparan sulfate 2-O-sulfotransferase 1, SDC3 = Syndecan 3.

Table 2.

Multivariate regression analysis. Specifics of multivariate COX regression analysis.

ID	Coef.	HR	HR.95L	HR.95H	P value
HS2ST1	−0.84277	0.430517	0.205691	0.901085	.025329
SDC3	−0.48457	0.615959	0.437241	0.867727	.005582

HS2ST1 = heparan sulfate 2-O-sulfotransferase 1, SDC3 = Syndecan 3.

Figure 4.

Risk factor plot. We ranked the risk values of the patients from lowest to highest in (A). Red dots represent death, green dots represent survival, and the survival status of each patient is shown in (B). The gene expression values of the 2 genes modeled in each case are shown in (C).

3.4. Survival analysis

Figure 5A and B show that the Kaplan–Meier survival curve is based on the gene expression of HS2ST1 and SDC3, and the prognosis of both gene high expression group is higher than that of the low gene expression group, but the difference is not statistically significant (P > .05). We also show the survival curves of some other glycolysis-related genes grouped based on their differential high and low expression (Fig. 5C–I). Furthermore, from the constructed Kaplan–Meier survival curve, a prognostic model for OS (Fig. 5J), the prognosis of the high-risk group was significantly lower than that of the low-risk group, and the difference was statistically significant (P < .05). This shows that there are practical implications of the prognostic model we constructed.

Figure 5.

Survival curves and ROC curves. (A–I) Survival curves of osteosarcoma constructed based on the high and low expression of genes. (J) survival curves constructed based on the prognostic model. (K) ROC diagnostic curve. ROC, receiver operating characteristic curve.

3.5. Calibration of the prognostic model of OS

Constructed ROC curves in Figure 5K reveal that the areas under the curve for predicting the probability of survival at 1, 2, and 3 years have AUC values of 0.811, 0.716, and 0.715, respectively, and all of which are higher than 50%. Nomogram-predicted calibration plot (Fig. 6C) illustrates that the predicted values of the prognosis (red line segment) significantly overlap the actual line segment. The column line graph (Fig. 6D) can help predict the survival probability of patients at 1, 2, and 3 years. We observed that the expression of HS2ST1 and SDC3 genes, for which the model was constructed, was higher in the high-risk group than in the low-risk group (Fig. 6A), and the difference was found statistically significant (P < .001). The 2-dimensional principal component analysis plot (Fig. 6B) showed all the patients in the high-risk group on the left side of the 0 scale of the PC1 coordinate axis, while those in the low-risk group on the right side of the 0 scale, with a clear distinction between the 2 groups.

Figure 6.

Validation plots of the prognostic model. (A) The gene expression of HS2ST1 and SDC3 in the high-risk group is higher than that in the low-risk group. (B) Patients in the high-risk group are located on the left side of the PC1 axis, while patients in the low-risk group are located on the right side of the PC1 axis. (C) Predicted prognosis at 3 years, and the predicted line segment (red) roughly overlaps with the actual line segment (gray). (D) Survival at 1, 2, and 3 years from patients can be predicted by gene expression of HS2ST1 and SDC3. HS2ST1 = heparan sulfate 2-O-sulfotransferase 1, SDC3 = Syndecan 3.

3.6. Immune cell typing analysis

To examine OS-associated immune cells and mine the expression profile more exhaustively, we used CIBERSORT software. We analyzed the immune cell composition of 88 OS samples (Fig. 7A); each sample had 22 immune cells and had only immune cells. Furthermore, we constructed a heat map showing a correlation between different immune cells and analyzed the correlation between both immune cells (Fig. 7B). Besides, we analyzed the immune cell expression of 88 OS cells based on the expression of HS2ST1 gene using CIBERSOTRT software for a thorough analysis of both genes employed to construct the model and observed a close association of HS2ST1 to T cells regulation that showed a negative correlation trend (Fig. 8A and B). Interestingly, immune cell analysis of SDC3 showed a significantly positive correlation to mast cells resting (Fig. 9A and B).

Figure 7.

Immune cell composition diagram. (A) Composition of the 22 immune cells for each osteosarcoma case we analyzed. (B) Relationship between each 2 immune cells, with red connections indicating positive correlation and green connections indicating negative correlation.

Figure 8.

Immune cell composition of osteosarcoma cases based on HS2ST1. (A) The immune cell composition of osteosarcoma patients based on HS2ST1 is shown, where Tregs is statistically significant (P value < .05). (B) HS2ST1 showed a negative correlation trend with Tregs (R = −0.24, P value = .026). HS2ST1 = heparan sulfate 2-O-sulfotransferase 1, Tregs = T cells regulatory.

Figure 9.

Immune cell composition of osteosarcoma cases based on SDC3. (A) The immune cell composition of osteosarcoma patients based on SDC3 is shown, where Mast cells resting is statistically significant (P value < .05). (B) SDC3 showed a positive correlation trend with Mast cells resting (R = 0.25, P value = .021). SDC3 = Syndecan 3.

3.7. Correlation analysis of modeled genes and drug sensitivity

As we can see from Figure 10, both HS2ST1 and SDC3, which were used to construct the prognostic model of OS, were significantly correlated with the drug sensitivity of multiple drugs. SDC3 showed a significant positive correlation with various drugs such as Encorafenib, Vemurafenib, Dabrafenib, Cobimetinib, and ARRY-162 (Cor > 0, P < .05). SDC3 also showed a negative correlation with Asparaginase, Nitrogen mustard, Oxaliplatin, and Ifosfamide (Cor < 0, P < .05). Similarly, HS2ST1 was also strongly correlated with various drugs such as Nelarabine, AZACITIDINE, and Fluorouracil (P < .05). If the gene is positively correlated with drug sensitivity, the stronger the gene expression, the more sensitive it is to the drug. If the gene is negatively correlated with drug sensitivity, the stronger the gene expression, the less sensitive it is to the drug. We provide a new reference for the use of gene-specific drugs for the treatment of OS.

Figure 10.

Analysis of genes and drug sensitivity in model construction. The graphs show the relationship between the sensitivity of each of SDC3 and HS2ST1 to multiple drugs. If there is a positive correlation, the stronger the gene expression, the greater the sensitivity to this drug. If there is a negative correlation, the stronger the gene expression, the weaker the sensitivity to the drug.

3.8. Immunohistochemical analysis

After immunohistochemical staining, the stained pathological sections were kept under an inverted microscope to examine their protein expression. The protein expression of HS2ST1 in OS (Fig. 11A1 and A2) was significantly higher than that in paraneoplastic tissues (Fig. 11B1 and B2). Besides, the protein expression of SDC3 in OS (Fig. 11C1 and C2) was also remarkably higher than in paraneoplastic tissues (Fig. 11D1 and D2). We could find that the positive rate of HS2ST1 was much higher in OS than in paraneoplastic tissues (Fig. 11E), and the difference was statistically significant (P < .05). Also, the positive rate of SDC3 was much higher in OS than in paraneoplastic tissue (Fig. 11F), and the difference was statistically significant (P < .05).

Figure 11.

Immunohistochemical analysis. (A1) and (aA) show the 100× magnification and 400× magnification of HS2ST1 expression in osteosarcoma, (B1) and (B2) show the 100× magnification and 400× magnification of HS2ST1 expression in paraneoplastic tissue. (C1) and (C2) indicate 100× magnification and 400× magnification plots of SDC3 expression in osteosarcoma; (D1) and (D2) indicate 100× magnification and 400× magnification plots of SDC3 expression in paraneoplastic tissues. (E) The immunohistochemical positivity statistics of HS2ST1. (F) The immunohistochemical positivity statistics of SDC3. “*” indicates P < .05, “*” indicates P < .01 “*” indicates P < .001. HS2ST1 = heparan sulfate 2-O-sulfotransferase 1, SDC3 = Syndecan 3.

4. Discussion

Here, the GO terms with enriched differentially expressed genes were mainly in glycosaminoglycan metabolic process, glycosaminoglycan biosynthetic process, aminoglycan biosynthetic process, etc. Glycosaminoglycans molecules have been expressed on all mammalian cells. They usually covalently bind to proteins to form proteoglycans and play a vital role in many physiological and pathological processes, such as cancer, bacterial, and viral infections.^[26] On the other hand, the KEGG pathway is mainly enriched in Glycosaminoglycan biosynthesis-heparan sulfate/heparin, Glycolysis/Gluconeogenesis, and insulin resistance. Hu et al^[27] showed that Ubiquitin Like with PHD And Ring Finger Domains 1, a gene that drives the silencing of tumor suppressor genes, is overexpressed in cancer, and that silencing Ubiquitin Like with PHD And Ring Finger Domains 1 significantly inhibits aerobic glycolysis in pancreatic cancer cells. Glucose transporter proteins and metabolic enzymes play an important role in the metastasis and progression of cancer, and metabolic changes and upregulation of glycolysis have been observed in many primary and metastatic cancers, and aerobic glycolysis is the most favorable mechanism of sugar metabolism in cancer cells, which is an evolutionary change.^[28] Accelerated aerobic glycolysis is not the primary result of dysfunctional or impaired mitochondria compensating for low ATP production per mole of glucose; in most tumors it is a component of (normoxia/hypoxia) hypoxia-inducible factor-1 overexpression, oncogene activation (cMyc, Ras) and the tumor microenvironment.^[29] Interestingly, it has also been found that hexokinase (HK) plays a key role in various biological processes such as glycolysis in tumor cells, that HK family genes are commonly genetically altered, and that the expression levels of the HK family correlate significantly with the activity of cancer marker-related pathways.^[30] On the other hand, many cancer cells are able to keep aerobic glycolysis at high levels of activation due to an irreversible mitochondrial oxidative phosphorylation deficiency.^[31] Adenine nucleotide translocases are a class of transporter proteins located in the inner mitochondrial membrane that act not only to link processes of cellular productivity and energy expenditure, but also to participate in the composition of the mitochondrial membrane permeability transition pore. Further adenine nucleotide translocases play an essential role in tumorigenesis, providing the basis for tumor control of oxidative phosphorylation, anabolic and glycolytic homeostasis, and control of cell death. It can provide the basis for tumor control of oxidative phosphorylation, anabolic and glycolytic homeostasis and control of cell death.^[32] This is consistent with our findings. In our study, by analyzing the role of glycolysis-related genes in OS, we found that HS2ST1 and SDC3 presented as significantly dysregulated in OS.

HS2ST1 is a protein-coding gene profoundly associated with hereditary spherocytosis type 3, and Overhydrated Hereditary Stomatocytosis diseases. Vijaya Kumar et al^[33] found overexpression of HS2ST1 and siRNA knockdown in breast cancer cell lines and documented that the invasiveness of HS2ST1 on breast cancer cells resulted from the compound effect of altered E-cadherin and EGFR expression, which led to altered mitogen-activated protein kinase and other signaling pathways. The 2-O sulfation of iduronic acid and 3-O sulfation of glucosamine in heparan sulfate (HS) are mediated by HS2ST and HS3ST sulfotransferases, respectively, and the 2 enzymes are found to be dysregulated in expression in a variety of cancers.^[34] Acetyl HS, which has multiple sulfation patterns, is one of the important regulators of cancer cells by interacting with multiple growth factors, while HS2ST1 was found to be downregulated at both mRNA and protein levels during granulocyte differentiation in SKM-1 leukemia cells, and HS (glucosamine) 3-O sulfotransferase 3A (HS3ST3A) in epithelial-mesenchymal transition of A549 lung cancer cells, while cell surface HS recognized by anti-HS antibodies was also altered in both cancer cell lines.^[35] This is consistent with our study. Our study suggests that HS2ST1 also presents as dysregulated in OS and its significant dysregulation may be a key factor contributing to the poor prognosis of OS. In summary, several researchers have established the role of HS2ST1 in cancer. Interestingly, we observed a strong association of HS2ST1 with Tregs in OS development, showing a remarkably negative trend. Tregs are a relatively small population of T cells in the body that play a key role in maintaining the balance of the body’s immune system. These cells are increased in almost all cancers, possibly to suppress the body’s immune system’s response to cancer cells.^[36] Surprisingly, Tregs penetrate deep into various tumor tissues and can promote tumor progression by suppressing anti-tumor immunity and supporting tumor evasion.^[37] In addition, several studies have revealed a strong association between Tregs and some tumors.^[38–42] Our results showed that HS2ST1 showed a significant negative correlation with T cells regulatory (Tregs), in other words, the high expression of HS2ST1 could significantly inhibit the expression of this immune cell. This mechanism may also contribute to the poor prognosis of OS. In addition, we found a close relationship between HS2ST1 and the drug sensitivity of several antitumor drugs, which provided a new reference for later investigators to use drugs to treat this malignant disease with poor prognosis.

SDC3, a protein-coding gene, is most closely associated with the Bardet-Biedl Syndrome 2 and Body Mass Index Quantitative Trait Locus on chromosome 11 disorders. SDC3 has been shown a reliable prognostic biomarker for breast cancer.^[43] Also, SDC3 reported in epithelial ovarian cancer can distinguish a malignant tumor from a benign tumor in epithelial ovarian cancer.^[44] Prostate cancer is the leading cause of cancer-related death in men, and Santos et al^[45] found that SDC3 was associated with greater tumor aggressiveness and poorer prognosis through RNA sequencing and using SDC 1-4 and SDCBP mRNA levels and patient survival data from the Cancer Genome Atlas and CamCAP databases. It has also been shown that TFPIα and SDC3 are closely related in vascular cells and cancer cells.^[46] It has also been shown that Circ-SCARB1 promotes renal cell carcinoma progression by isolating miR-510-5p and indirectly upregulating SDC3 expression.^[47] Overexpression of SDC3 was closely associated with poor prognosis of renal cell carcinoma, and its expression enhanced the invasive ability of cancer cells.^[48] This is consistent with our findings. Our study suggests that SDC3 presents as a significant dysregulation in OS, and its possible involvement in the progression of OS together with HS2ST1. Therefore, SDC3 plays a pivotal role in cancer development. Our findings also showed a significant positive trend of correlation between SDC3 and mast cells resting. Mast cells resting is found to be closely related to soft tissue sarcoma.^[49] Mast cells resting are resident cells that can participate in tissue homeostasis and finely regulate not only the immune response but also the pathogenesis of diseases, including cancer.^[50] This would also suggest that the high expression of SDC3 in OS may activate the expression of Mast cells resting, thus promoting the development of OS cancer cells. In addition, we found a close relationship between SDC1 and the drug sensitivity of several antitumor drugs, which provided a new reference for later investigators to use drugs to treat this malignant disease with poor prognosis.

Overall, we have used advanced bioinformatics techniques to deeply profile the prognostic biomarkers of OS and analyzed their correlation with immune cells, thereby suggesting a possibility for immunotherapy to treat OS. We also analyzed the relationship between these 2 genes and drug sensitivity, providing a new reference point for the gene-based use of drugs in the treatment of OS. Finally, we analyzed the protein expression of HS2ST1 and SDC3 in OS and paraneoplastic tissues by immunohistochemistry, further validating the results and reliability of our analysis.

As with all other studies, our study has certain limitations. First, the inadequacy of the sample size, which we used, was far from adequate relative to the large sample size. Second, we did not use enough laboratory manipulations to test our results.

5. Conclusion

HS2ST1 and SDC3 are significantly dysregulated in OS, and the prognostic models constructed based on these 2 genes have much lower survival rates in the high-risk group than in the low-risk group. HS2ST1 and SDC3 can be used as glycolytic and immune-related prognostic biomarkers in OS.

Acknowledgments

The authors are grateful to Dr Xin Li Zhan (Spine and Osteopathy Ward, The First Affiliated Hospital of Guangxi Medical University) for his kindly assistance in all present study stages.

Author contributions

GY, JJ, CL, and XZ designed the study. RY analyze the data. ZL digital visualization. GY wrote and revised the manuscript. CL and XZ revised the manuscript. All authors read and approved the final manuscript.

Guozhi Yang, Chong Liu, Xinli Zhan.

Guozhi Yang, Chong Liu, Xinli Zhan.

Guozhi Yang, Jie Jiang, Ruifeng Yin, Feng Gao, Chong Liu, Xinli Zhan.

Guozhi Yang, Ruifeng Yin, Chong Liu, Xinli Zhan.

Guozhi Yang, Ruifeng Yin, Zhian Li, Chong Liu, Xinli Zhan.

Guozhi Yang, Jie Jiang, Zhian Li, Chong Liu, Xinli Zhan.

Guozhi Yang, Zhian Li, Lei Li, Chong Liu, Xinli Zhan.

Guozhi Yang, Jie Jiang, Zhian Li, Lei Li, Chong Liu, Xinli Zhan.

Guozhi Yang, Lei Li, Chong Liu, Xinli Zhan.

Guozhi Yang, Lei Li, Chong Liu, Xinli Zhan.

Guozhi Yang, Feng Gao, Chong Liu, Xinli Zhan.

Guozhi Yang, Feng Gao, Chong Liu, Xinli Zhan.

Guozhi Yang, Chong Liu, Xinli Zhan.

Guozhi Yang, Chong Liu, Xinli Zhan.