Identification of differentially expressed TRGs

According to the findings in Fig. 2A, the GSEA analysis indicated a significant enrichment of the TGF-β pathway in the tumor group. By employing the GSVA algorithm, we calculated the TGF-β pathway gene set score for each sample. It is important to highlight that the oral cancer group exhibited a significantly higher score for the TGF-β pathway compared to the normal group, as shown in Fig. 2B (p < 0.01). These results imply a potential essential role of the TGF-β pathway gene set in the advancement of oral cancer, underscoring the need for further investigation and analysis of this particular gene set. After conducting a comparative analysis between the groups with oral cancer and normal individuals, it was discovered that a total of 16 TRGs exhibited differential expression. These genes were all observed to be upregulated in the tumor group. Specifically, the upregulated genes included ID3, TGFB3, LTBP2, PMEPA1, BMP2, SERPINE1, TGFB1, TGIF1, SMURF1, SKIL, ACVR1, CDK9, RAB31, TGFBR1, SLC20A1, and SMURF2 (Fig. 2C, D).

Fig. 2

Screening of differentially expressed TRGs. A GSEA result showed a significant enrichment of the TGF-β pathway in the tumor group. B A box plot was employed to visually illustrate the TGF-β pathway score comparison between the normal group and the oral cancer group. **p < 0.01. C The volcano plot depicting the expression level of TRGs. D A heatmap was employed to visually display the expression levels of TRGs in both the normal and oral cancer groups

Enrichment analysis of differentially expressed TRGs

As depicted in Fig. 3A, the results from the GO analysis indicated a notable enrichment of these TRGs in the transforming growth factor beta receptor signaling pathway, transmembrane receptor protein serine/threonine kinase signaling pathway, SMAD binding, cellular response to transforming growth factor beta stimulus, etc. KEGG analysis revealed that these TRGs were significantly enriched in the TGF-beta signaling pathway, hippo signaling pathway, cytokine-cytokine receptor interaction, cellular senescence, etc. (Fig. 3B).

Fig. 3

Enrichment analysis of differentially expressed TRGs. Functional annotation analyses were performed using GO enrichment analysis (A) and KEGG enrichment analysis (B)

Establishment of a prognostic signature related to TRGs

The LASSO regression analysis was performed using the previously mentioned 16 TRGs. Screening of the data revealed eight TRGs (TGFB3, LTBP2, BMP2, TGFB1, ACVR1, CDK9, SLC20A1, and SMURF2) with non-zero coefficients, indicating their association with the prognosis of oral cancer patients (Fig. 4A, B). The Kaplan-Meier curve demonstrated that patients with OSCC who were classified as high risk experienced significantly reduced overall survival compared to those classified as low risk (Fig. 4C). Furthermore, it was observed that a significant proportion of TRGs were upregulated in the subgroup with a high risk (Fig. 4D).

Fig. 4

Establishment of a prognostic signature related to TRGs. A, B LASSO variable trajectory plots were analyzed to identify non-zero variables. C Comparison of survival outcomes between OSCC patients classified as high risk and low risk in the TCGA-OSCC dataset. D A heatmap was employed to visually display the expression levels of TRGs in both the low-risk and high-risk subgroups. **p < 0.01; ***p < 0.001

Immuno-infiltration analysis of the TRG-related signature

By utilizing ssGSEA, we investigated the association between the risk score and immune infiltration and found that patients with a lower risk score showed increased infiltration of immune cells in comparison to those with a higher risk score (Fig. 5A, B). By conducting ESTIMATE algorithm, we demonstrated that the low-risk group had higher ESTIMATE score, stromal score, and immune score, whereas the low-risk group had lower tumor purity compared to the high-risk group (Fig. 5C–F).

Fig. 5

Immune infiltration analysis. We utilized the heatmap (A) and box plot (B) to visually illustrate the enrichment fraction of infiltrating cells in both high-risk and low-risk subgroups. Variations in the ESTIMATEScore (C), ImmuneScore (D), StromalScore (E), and tumor purity (F) among patients with high-risk and low-risk statuses. **p < 0.01; ***p < 0.001

GSEA

GSEA results revealed that DNA repair, p53 pathway, glycolysis, G2M checkpoint, PI3K-AKT MTOR signaling, TGF beta signaling, fatty acid metabolism, and apoptosis were significantly enriched in the high-risk subgroup (Fig. 6).

Fig. 6

GSEA was conducted to investigate the potential pathways associated with the low-risk and high-risk subgroups

SMURF2 emerged as a significant prognostic indicator for patients diagnosed with oral cancer.

In OSCC patients, a multivariate regression analysis was conducted on the eight TRGs, revealing that SMURF2 emerged as an independent prognostic factor (Table 1). The results from Kaplan-Meier analysis indicated a significant correlation between increased SMURF2 expression and decreased overall survival in patients with oral squamous cell carcinoma (Fig. 7A and Figure S1). The findings from the TCGA-OSCC dataset (Fig. 7B) and analysis conducted on TNMplot.com (Fig. 7C, D) revealed that the mRNA expression level of SMURF2 was significantly higher in the tumor group in comparison to the normal group (p < 0.001 or p < 0.01). In addition, by utilizing the HPA database (Fig. 7E), we have also discovered that the protein expression level of SMURF2 exhibited a marked increase in the tumor group as compared to the normal group. As shown in Figure S2, the comparative analysis of SMURF2 gene expression across various cancer types indicated differential expression patterns between normal and tumor tissues. In the case of adrenocortical carcinoma (ACC), the expression of SMURF2 was significantly lower in tumor samples as compared to normal tissue. Similar patterns of reduced SMURF2 expression in tumor samples were observed in kidney chromophobe (KICH), lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), ovarian serous cystadenocarcinoma (OV), skin cutaneous melanoma (SKCM), testicular germ cell tumors (TGCT), uterine corpus endometrial carcinoma (UCEC), and uterine carcinosarcoma (UCS), all showing statistical significance (p < 0.05). Conversely, for cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), esophageal carcinoma (ESCA), glioblastoma multiforme (GBM), head and neck squamous cell carcinoma (HNSC), kidney renal papillary cell carcinoma (KIRP), acute myeloid leukemia (LAML), brain lower grade glioma (LGG), liver hepatocellular carcinoma (LIHC), pancreatic adenocarcinoma (PAAD), rectum adenocarcinoma (READ), stomach adenocarcinoma (STAD), and thymoma (THYM), tumor tissues demonstrated significantly higher expression levels of SMURF2 when compared to normal tissues (p < 0.05).

Table 1 Results of multivariate regression analysisFig. 7

SMURF2 was a significant predictor of prognosis in patients with oral cancer. A Comparison of survival outcomes between OSCC patients classified as high-SMURF2 and low-SMURF2 in the TCGA-OSCC dataset. The expression of SMURF2 was examined in both the TCGA-OSCC dataset (B) and through analysis on the TNMplot.com platform (C, D) to compare the levels between normal and tumor groups. E The protein expression level of SMURF2 in both normal and tumor tissue

Development of nomogram

We have developed a nomogram to forecast the overall survival rates at 1, 3, and 5 years for patients diagnosed with oral cancer. The nomogram, depicted in Fig. 8A, includes the expression of the SMURF2 gene. The calibration curve, shown in Fig. 8B, demonstrates a strong correlation between the predicted outcomes and the actual results observed in oral cancer patients. Furthermore, Fig. 8C highlights the high diagnostic accuracy of the SMURF2 gene, with an area under the curve (AUC) of 0.838. The findings of DCA demonstrated that SMURF2 exhibited a favorable overall benefit and possessed the ability to forecast the overall survival of individuals diagnosed with oral cancer, both in the short and long term (Fig. 8D–F).

Fig. 8

Development and assessment of the nomogram. A The nomogram provides the ability to assess the likelihood of survival at 1-year, 3-year, and 5-year time points for individuals with oral cancer. B The accuracy of the nomogram in predicting oral cancer patient outcomes is evident in its well-calibrated curve. C The diagnostic value of SMURF2 as a marker for oral cancer is shown through the ROC curve analysis. D–F Survival rates of oral cancer patients were analyzed using DCA curves for 1, 3, and 5 years, focusing on the SMURF2 gene

Single-gene GSEA

Single-gene GSEA results revealed that PI3K-AKT MTOR signaling, glycolysis, G2M checkpoint, TGF beta signaling, hypoxia, WNT beta catenin signaling, DNA repair, IL2 STAT5 signaling, p53 pathway, inflammatory response, IL6-JAK-STAT3 signaling, and TNFA signaling via NFKB were significantly enriched in high-SMURF2 subgroup (Fig. 9).

Fig. 9

Single-gene GSEA was conducted to investigate the potential pathways associated with the low-SMURF2 and high-SMURF2 subgroups

Investigation of the immune cell infiltration patterns within the subgroups of SMURF2

As shown in Fig. 10A, B, the ssGSEA results demonstrated significantly different levels of immune cell infiltration between the high and low SMURF2 subgroups. Correlation analysis demonstrated a positive correlation between the SMURF2 gene and the infiltration level of Th2 cells, T helper cells, NK cells, Tcm, eosinophils, Tem, macrophages, and Tgd. On the other hand, a negative correlation was observed between the SMURF2 gene and the infiltration level of Th17 cells, cytotoxic cells, pDC, NK CD56dim cells, T cells, and B cells (Fig. 10C).

Fig. 10

Investigation of the immune cell infiltration patterns within the subgroups of SMURF2. The visual representation of the 24 immune cell types in both the low-SMURF2 and high-SMURF2 groups was presented using a heatmap (A) and box plot (B). C To illustrate the correlation between SMURF2 expression and immune cell infiltration levels, a Lollipop plot was employed. *p < 0.05; **p < 0.01; ***p < 0.001

Validation of SMURF2 expression by qRT-PCR analysis

As shown in Fig. 11, the qRT-PCR findings demonstrated that the expression level of SMURF2 was increased in SCC15, SCC9, SCC4, HSC4, and CAL27 cells compared to HOEC cells (p < 0.05 or p < 0.001). These results align with our bioinformatics analyses.

Fig. 11

Cell experiments were conducted to validate the expression of SMURF2. The expression level of SMURF2 was compared between various oral cancer cell lines (SCC15, SCC9, SCC4, HSC4, and CAL27) and a human oral gingival epithelial cell line (HOEC cells). *p < 0.05; ***p < 0.001