Standardized pan-cancer data were retrieved from databases including The Cancer Genome Atlas (TCGA), Therapeutically Applicable Research to Generate Effective Treatments (TARGET) and Genotype-Tissue Expression using the University of California Santa Cruz (UCSC) genome browser database (https://xenabrowser.net/datapages/) (8,9). The abbreviations used for 34 types of cancer are presented in Table SI. The data on the SLC17A9 expression pattern in all samples were extracted and plotted using the R software. However, the data on SLC17A9 expression levels in corresponding healthy tissues for certain types of cancer from TCGA were missing. Thus, data on the SLC17A9 expression levels in these types of cancer and corresponding healthy tissues, including adrenocortical carcinoma (ACC), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), acute myeloid leukemia (LAML), low grade glioma (LGG), ovarian serous cystadenocarcinoma, skin cutaneous melanoma (SKCM), testicular germ cell tumors (TGCT), thymoma (THYM), and uterine carcinosarcoma (UCS), were obtained from the Gene Expression Profiling Interactive Analysis (GEPIA; http://gepia.cancer-pku.cn/) database (10).

To determine the association between SLC17A9 expression and survival outcomes of various types of cancer, the OS, progression-free survival (PFS) and disease-specific survival (DSS) data of patients were obtained from the TCGA and TARGET cohorts. TCGA data including expression data and clinical data were obtained from the University of California, Santa Cruz (UCSC) database, which is an open, pubic database (https://xenabrowser.net/datapages/). In the UCSC database, not all tumor types contained the three survival times (OS, PFS, and DSS), some cancer types only contained OS or PFS. Survival analysis was performed using the Kaplan-Meier (KM) method with the median of SLC17A9 expression as the cut-off value and the survival curves were plotted using the survival package in R. In addition, only when P was less than 0.05 can the survival curves be graphed. Next, a univariate Cox regression analysis was performed using the forestplot R package to determine the association between SLC17A9 expression levels and survival. In addition, the association between SLC17A9 expression levels and the survival of patients in the immunotherapy cohort was determined using KM plotter (https://kmplot.com/analysis/) (11). The detailed steps were as follows: i) Enter the KM plotter website and select the ‘start KM plotter for immunotherapy’ button; ii) enter the gene symbol ‘SLC17A9’ and select the cut-off value ‘median’; and iii) select survival ‘OS or PFS’, select follow-up threshold ‘all’, select Anti-PD-L1 treatment ‘all’, select tumor type ‘all, bladder...’.

Stemness refers to the self-renewal and dedifferentiation properties of cells, which aid in the progression and invasion of cancer cells, thus resulting in poor prognosis of the patient (12). The two independent stemness indices are the mRNA expression-based stemness index (RNAss), which demonstrates gene expression, and the DNA methylation-based stemness index (DNAss), which reflects epigenetic features. The stemness scores of the patients were obtained from the TCGA cohort based on the UCSC database. The correlation between SLC17A9 expression levels and the stemness indices was determined by Spearman correlation analysis.

The tumor microenvironment (TME) serves a crucial role in cancer progression. The proportion of stromal and immune cells in the TME of patients was determined from the TCGA cohort by calculating the estimation of stromal and immune cells in malignant tumor tissues using expression data (ESTIMATE) score (13) determined with the ESTIMATE algorithm (https://bioinformatics.mdanderson.org/public-software/estimate/). In addition, the cell-type identification by estimating relative subsets of RNA transcripts (CIBERSORT) tool (https://cibersortx.stanford.edu/) was used to determine the infiltration status of tumor-infiltrating immune cells (TIICs) in 34 types of cancer (14). Subsequently, the effect of SLC17A9 expression on the TME and TIICs was determined.

To predict potential drugs targeting SLC17A9, the correlation between SLC17A9 expression levels and drug sensitivity was determined by Pearson's correlation coefficient. Furthermore, the CellMiner database (https://discover.nci.nih.gov/cellminer/) was examined to obtain data on drug activity (15).

The clusterProfiler R package (16) was employed to perform the GO and KEGG pathway enrichment analyses to determine the functions and pathways enriched by SLC17A9. Next, the GSVA R package (17) was used to estimate the variations in key gene sets in patients with OSS.

OSS cells were obtained from The Cell Bank of Type Culture Collection of The Chinese Academy of Sciences. U2OS, Saos2, MG63 and HOS cells were cultured in Dulbecco's Modified Eagle's Medium supplemented with 10% fetal bovine serum (both from Biological Industries), and 143 B cells were cultured in Modified Eagle's Medium (Biological Industries) supplemented with 10% fetal bovine serum. All cells were incubated at 37°C in a humidified atmosphere containing 5% CO2. GV492 was used as the vector for overexpression and empty plasmid, and SLC17A9 overexpression (OE) and empty (NC) vectors were constructed and provided by Shanghai GeneChem Co., Ltd. Small interfering RNAs (siRNAs) targeting SLC17A9 (si-1, stB0012921A and si-2, stB0012921B) and negative control siRNA (si-nc, siN0000001-1-5) were designed and synthesized by Guangzhou RiboBio Co., Ltd. The si-1 and si-2 target sequences are presented in Table SII. These vectors or siRNAs were separately transfected into cells with the aid of Lipofectamine 3000 (Invitrogen; Thermo Fisher Scientific, Inc.) or riboFECT CP Transfection Kit (cat. no. C10511-05; Guangzhou RiboBio Co., Ltd). HOS and MG63 (1×10⁶/well) cells were seeded into a 6-well plate. After 24 h, vectors and Lipofectamine 3000 (or siRNAs and riboFECT CP reagent) were mixed and left at room temperature for 15 min, then the mixture was added to each well of the 6-well plate. MG63 cells were transfected with 5 µg vector in each well. HOS cells were transfected with 100 nM siRNA per well. The transfected cells were incubated for 48 h at 37°C in a humidified atmosphere containing 5% CO2, and then used in subsequent experiments.

RT-qPCR and WB were performed as described previously (18). RT-qPCR reagents such as AG RNAex Pro Reagent, SYBR Green Premix Pro Taq HS qPCR kit and the Evo M-MLV RT Premix kit were purchased from Accurate Biology. The sequences of the primers used are presented in Table SII. The reagents used to perform WB, including RIPA buffer and BeyoECL Plus kit, were purchased from Beyotime Institute of Biotechnology. Furthermore, anti-GAPDH (1:1,000; cat. no. 10494-1-AP) and anti-SLC17A9 (1:500; cat. no. 26731-1-AP) antibodies were purchased from Wuhan Sanying Biotechnology, and the HRP-conjugated Affinipure Goat Anti-Rabbit IgG antibody (1:1,000; cat. no. SA00001-2) was purchased from Proteintech Group, Inc (Wuhan Sanying Biotechnology).

The proliferation of cells was determined using the CCK-8 assay (Beyotime Institute of Biotechnology) following the manufacturer's protocols. For this assay, 2×10³ cells/well were seeded and cultured in a 96-well plate at 37°C for 0, 24, 48 and 72 h. For the colony-formation assay, 1,000 cells/well were seeded and incubated in a 6-well plate and incubated for 14 days, with the media being changed every 2 days. Next, the colonies were stained at room temperature with 1% crystal violet for 20 min. The colonies that had formed (colonies with ≥50 cells) were manually counted under a microscope.

Bioinformatics and statistical analyses, such as the normalization and transformation of RNA sequencing data, and the correlation, survival, CIBERSORT, ESTIMATE, GSVA, GO and KEGG analyses, were performed using the R software (version 4.0.0; http://www.r-project.org/). Furthermore, the coxph algorithm of the survival R package was used to perform univariate and multivariate Cox regression analyses. Pearson correlation analysis was performed to determine the correlation between SLC17A9 expression levels and the activity of the drugs from the CellMiner database. All quantitative data from in vitro experiments are expressed as the mean ± standard deviation of three independent experiments. One-way analysis of variance followed by Tukey's post hoc test was performed using the GraphPad Prism 8.0 Software (Dotmatics) to determine differences across groups. A two-tailed P<0.05 was considered to indicate a statistically significant difference.

The differences in SLC17A9 expression profiles between pan-cancer tumor and normal samples were analyzed after obtaining data of the SLC17A9 expression profiles for pan-cancer using publicly available data. A significant increase in SLC17A9 expression levels was observed in 15 types of cancer (Fig. 1A and B), including LUAD, COAD, BRCA, KIRP, DLBC, head and neck squamous cell carcinoma (HNSC), LAML, KIRC, LIHC, BLCA, READ, STAD, LUSC, THCA and UCEC. However, a significant decrease in SLC17A9 expression levels was observed in the tissues of patients with ACC (Fig. 1B).

Univariate Cox regression and survival analyses were performed to determine the association between SLC17A9 expression levels and the prognosis of patients in 34 types of cancer. The univariate Cox regression analysis demonstrated that SLC17A9 expression was an independent prognostic factor associated with the OS of patients with various types of cancer, such as OSS, ACC, LGG, KIRC, MESO, KIRP, SKCM and UVM (Fig. 2A). Furthermore, the KM analysis demonstrated that patients with high SLC17A9 expression and types of cancer such as KIRC, LGG, OSS or UVM had a reduced OS duration compared with patients with low SLC17A9 expression. On the contrary, the OS duration of patients with high SLC17A9 expression levels and BRCA was increased compared with that of patients with low SLC17A9 expression levels (Fig. S1).

The SLC17A9 expression levels were an independent risk factor for DSS in patients with types of cancer such as KIRC, ACC, KIRP, LGG, MESO, LIHC, SKCM, LUAD, STAD and UVM (Fig. 2B). KM analysis demonstrated that the DSS duration of patients with high SLC17A9 expression levels and KIRC, LGG or UVM was reduced compared with that of patients with low SLC17A9 expression levels. However, the DSS duration of patients with high SLC17A9 expression levels and LUAD was increased compared with that of patients with low SLC17A9 expression levels (Fig. S1). Forest plots demonstrated a significant association between SLC17A9 expression levels and the PFS of patients with glioblastoma (GBM), LGG, KIRC, LUAD, PRAD, THYM, STAD and UVM (Fig. 2C). In addition, KM analysis demonstrated that the PFS duration of patients with high SLC17A9 expression levels and KIRC, LGG, PRAD or UVM was reduced compared with that of patients with low SLC17A9 expression levels (Fig. S1). The DSS and PFS of patients with high SLC17A9 expression levels and LUAD were increased compared with those of patients with low SLC17A9 expression levels.

The association between the SLC17A9 expression levels and the prognosis of patients from the immunotherapy cohort was also investigated using the KM plotter web-based tool. The results revealed that in several types of cancer, including bladder cancer, esophaseal adenocarcinoma, GBM, hepatocellular carcinoma, melanoma, HNSC, non-small cell lung cancer, non-squamous lung carcinoma and urothelial cancer, the OS of patients with high SLC17A9 expression levels in the antibody directed against programmed cell death-1 ligand 1 (anti-PD-L1) therapy cohort was increased (Fig. S2A). Furthermore, a significant association was observed between the SLC17A9 expression levels and the OS of patients with bladder carcinoma (Fig. S2B) in the anti-PD-L1 therapy cohort. Thus, SLC17A9 may be a potential prognostic marker for various types of cancer.

The correlation between the stemness indices and the SLC17A9 expression levels exhibited differences among various cancer types. A negative correlation was revealed between SLC17A9 expression levels and the RNAss stemness index in various types of cancer, such as ACC, CESC, GBM, BLCA, HNSC, KICH, LUAD, KIRP, KIRC, LUSC, PCPG, LGG, PRAD, TGCT, SKCM and THCA. However, a positive correlation was observed between SLC17A9 expression levels and the RNAss stemness index in other types of cancer, such as PAAD and STAD (Fig. 3A). In addition, the SLC17A9 expression level was negatively correlated with the DNAss stemness index in BLCA, CESC, LAML and TGCT, and positively correlated in other types of cancer, such as HNSC, BRCA, KIRC, LGG, MESO, KIRP, PAAD, THCA, PRAD, SARC, PCPG, STAD and UVM (Fig. 3A).

The ESTIMATE algorithm was used to calculate the stromal and immune scores of patients. A positive correlation was observed between the SLC17A9 expression levels and the immune score of patients with BRCA, KIRC, GBM, ACC, HNSC, ESCA, UCEC, KIRP, LGG, KICH, SKCM, LUAD, BLCA, MESO, OSS, PCPG, SARC, LUSC, TGCT, PRAD, THYM, THCA and UVM. However, a negative correlation was observed between the SLC17A9 expression level and the immune score of patients with COAD, LAML, PAAD and STAD (Fig. 3B). Furthermore, the results revealed a positive correlation between the SLC17A9 expression level and the stromal scores of patients with ACC, BLCA, ESCA, ESCA, KICH, GBM, HNSC, KIRP, LGG, LUSC, MESO, PRAD, SKCM, TGCT, THYM, SARC, THCA and UCEC. However, a negative correlation was observed between the SLC17A9 expression level and the stromal scores of patients with LAML, OSS, PAAD and STAD (Fig. 3B). Taken together, these results suggest a significant correlation between the SLC17A9 expression level and stemness indices such as RNAss and DNAss, as well as the immune and stromal scores of patients with BLCA, HNSC, LGG, PAAD, TGCT, PRAD, KIRP, STAD and THCA.

Next, whether there was a correlation between SLC17A9 expression and the levels of 22 types of TIIC was determined. The results demonstrated a positive correlation between SLC17A9 expression and nine TIICs, including plasma cells, naïve and memory B cells, memory-activated CD4 T cells, resting natural killer (NK) cells, regulatory T (Treg) cells, M0 macrophages, eosinophils and neutrophils. On the contrary, a negative correlation was observed between SLC17A9 expression levels and the levels of other types of immune cell, such as naïve CD4 T cells, activated NK cells, T follicular helper (Tfh) cells, monocytes, M1 and M2 macrophages, resting and activated dendritic cells, and activated mast cells (Fig. 3C). Furthermore, differences in TIIC levels were observed among patients with different types of cancer. In patients with ACC or DLBC, there was a strong positive correlation between the SLC17A9 expression levels and immune cells such as neutrophils and resting NK cells, respectively. Furthermore, SLC17A9 expression levels were revealed to be positively correlated with resting memory CD4 T cells in ESCA and to be negatively correlated with monocytes in LAML. In patients with PRAD, THYM or UVM, there was a positive correlation between the SLC17A9 expression levels and immune cells such as Treg cells, plasma cells and M1 macrophages, respectively (Fig. 3C). However, no significant correlation was observed between SLC17A9 expression levels and the infiltration of 22 immune cell types in patients with BLCA or UCS (Fig. 3C). These results indicate a correlation between the SLC17A9 expression level and the degree of immune-cell infiltration in different types of cancer.

In the dataset GSE16088, an increase in SLC17A9 expression levels was observed in the tissues of patients with OSS compared with those in healthy bone tissues (Fig. 4A). Furthermore, in the TCGA-TARGET dataset, the OS (Figs. 4B and S1) and relapse-free survival (Fig. 4B) of patients with high SLC17A9 expression levels were poor compared with those of patients with low SLC17A9 expression levels. Univariate and multivariate Cox regression analyses of the TCGA-TARGET dataset demonstrated that SLC17A9 may be an independent risk factor for the OS of patients with OSS (Fig. 4C).

Next, RT-qPCR was performed to determine SLC17A9 expression levels in several OSS cell lines (Fig. 4D). HOS and MG63 cells were used for the subsequent experiments since SLC17A9 levels were high in HOS and low in MG63 cells. HOS cells were transfected with siRNAs (si-nc, si-1 and si-2) and MG63 cells were transfected with SLC17A9-NC or SLC17A9-OE vectors. Finally, RT-qPCR and WB analyses were performed to confirm the knockdown and overexpression of SLC17A9 in HOS and MG63 cells, respectively (Fig. 4E). The results of the CCK-8 assay revealed a decrease in the proliferation capacity of SLC17A9-knockdown HOS cells (si-1 or si-2) compared with cells transfected with si-NC. In addition, there was an increase in the proliferation of MG63 cells transfected with the SLC17A9-OE vector compared with the cells transfected with SLC17A9-NC (Fig. 4F). The colony-formation assays indicated a considerable reduction in the colony number of HOS cells with SLC17A9 knockdown, while there was a substantial increase in the colony numbers of MG63 cells overexpressing SLC17A9 (Fig. 4G). These results indicate that SLC17A9 has a role to enhance the proliferation and viability of OSS cells. Thus, SLC17A9 may be significantly involved in enhancing OSS progression.

Next, GO and KEGG pathway enrichment analyses were performed to determine the functions and pathways enriched by SLC17A9 in OSS. Fig. 5A reveals that SLC17A9 was enriched in biological processes such as ossification, connective tissue and cartilage development, the bone morphogenetic protein signaling pathway and odontogenesis. In addition, SLC17A9 was enriched in cellular component terms such as the collagen-containing extracellular matrix. The molecular function terms enriched by SLC17A9 were extracellular matrix structural constituents, receptor serine/threonine kinase and TGF-β receptor binding, growth factors activity, racemase and epimerase activity. The KEGG pathway enrichment analysis demonstrated that SLC17A9 was enriched in the MAPK, Hippo, focal adhesion and TGF-β signaling pathways (Fig. 5B). Finally, GSVA was performed to compare the pathways enriched in patients with OSS with high or low SLC17A9 expression levels. The GSVA results revealed significant differences in pathways such as circadian rhythm in mammals, starch and sucrose metabolism, ABC transporters, glycine serine/threonine, histidine, phenylalanine, sulfur, glyoxylate and dicarboxylate metabolism and Vibrio cholera infection between the high and low expression groups (Fig. 5C).

Data on 263 anticancer drugs were retrieved from the CellMiner database. Next, Pearson correlation analysis was performed to determine the correlation between SLC17A9 expression levels and the activity of these drugs in NCI-60 cells (15). The scatterplot in Fig. 6 revealed the correlation analysis results: A positive correlation was observed between SLC17A9 expression levels and the sensitivity to drugs such as vorinostat, asparaginase, chelerythrine, hypothemycin, PX-316, entinostat, acrichine, LDK-378 and imexon. A negative correlation was observed between SLC17A9 expression levels and the sensitivity to drugs including ibrutinib, afatinib and sonidegib.