3.1 FUTs and gene mutation analysis

We utilized TCGA patient data to study CNVs in FUTs as part of our analysis. As demonstrated by a CNV distribution pie chart, heterozygous amplifications and deletions constituted the most prevalent forms of CNVs identified in patients (Fig. 1A).

Fig. 1

Copy number variation (CNV) distribution, single nucleotide variation (SNV) frequency and methylation distribution of FUTs in 33 tumors. A CNV pie chart showing the combined heterozygous/homozygous CNVs of each gene in each cancer. A pie chart representing the proportions of different types of CNVs of one gene in one cancer, where different colors represent different types of CNVs. Hete Amp = heterozygous amplification; Hete Del = heterozygous deletion; Homo Amp = homozygous amplification; Homo Del = homozygous deletion; None = no CNV. B SNV oncoplot. An oncoplot showing the mutation distribution of FUTs and a classification of SNV types. C Mutation frequency of FUTs. The numbers represent the number of samples that have the corresponding mutated gene for a given cancer. A ‘0’ indicates that there was no mutation in the gene coding region, and no number indicates that there was no mutation in any region of the gene. D Differential methylation in FUTs between tumor and normal samples in each cancer type. Blue indicates decreased methylation in tumors, and red indicates increased methylation in tumors; the darker the color is, the larger the difference in methylation level

In addition, an examination of single nucleotide polymorphisms (SNPs) in FUTs was undertaken to assess the frequency and variants of genes within each tumor subtype. Patients with UCEC, READ, LUSC, STAD, LUAD, COAD, SKCM, and BLCA exhibited a frequency range of 0% to 51% for SNVs in FUTs, as illustrated in Fig. 1B. An additional finding indicated that SNVs were present in regulators of 89.13% (623/699) of patients (Fig. 1C). Of these SNVs, SNP types with the highest prevalence among patients were missense mutations. Particularly, a high proportion of these mutations were accounted for by the top 10 mutated genes, namely FUT9, FUT8, FUT3, FUT5, FUT6, POFUT2, FUT10, FUT1, FUT2, and POFUT1, with frequencies varying from 8 to 23%.

3.2 Analysis of FUT-related differential methylation across various types of cancer

We investigated the epigenetic regulation of FUTs by examining the methylation state of these genes. Significant patient heterogeneity was observed in the FUTs methylation status (Fig. 1D). Hypermethylation of FUTs appears to be more prevalent in PRAD and BRCA, according to our findings. Conversely, an elevation in hypomethylation in FUTs was observed among patients with KIRP, BLCA, THCA, KIRC, LUSC, UCEC, LIHC, PAAD, and LUAD. In most cancers, genes such as FUT9 and FUT1 exhibited excessive methylation (FDR ≤ 0.05, Fig. 1D). Additionally, FUT3, FUT6, FUT5, FUT8, FUT7,and FUT2 exhibited reduced methylation in most cancers (FDR ≤ 0.05, Fig. 1D).

3.3 Differential expression of FUTs across cancers and their impact on pathway activity and prognosis

FUT expression variations among cancer patients were investigated. Differences in FUT expression were found to be significant among patients diagnosed with the following solid tumor types: THCA, LIHC, HNSC, LUAD, KIRC, KICH, COAD, STAD, PRAD, LUSC, KIRP, and BRCA (FDR ≤ 0.05, Fig. 2A). An analysis BLCA and ESCA individuals did not reveal a statistically significant variation in FUT expression (FDR ≤ 0.05, Fig. 2A).

Fig. 2

Differential expression of FUTs across cancers and their impact on pathway activity and prognosis. A The mRNA differences between normal samples and tumor samples. B The combined percentage of the effect of FUTs on pathway activity. C Differences in survival between patients with high and low gene expression. Red points indicate poor survival in the high-expression group, and blue points indicate poor survival in the low-expression group. The size of the point represents the statistical significance, where the larger the dot size is, the greater the statistical significance

The critical involvement of FUTs in cancer-associated pathways was established through pathway activity analysis. The signaling pathways, hormone-related pathways including ER and AR, EMT, the cell cycle, programmed cell death, PI3K/AKT, RAS/MAPK, RTK, response to DNA damage, and TSC/mTOR were among the pathways involved (Fig. 2B).

Furthermore, a strong link was observed between the FUTs' expression and the survival status of the patients (PFS, OS, DSS, and DFI) (Cox p < 0.05, Fig. 2C). These data indicated that aberrant FUT expression might play a role in the occurrence of tumors. Additionally, POFUT2 overexpression was negatively correlated with OS in 10 cancer types, DFS in 4 cancer types, DSS in 11 cancer types, and PFS in 9 cancer types (Cox P < 0.05, Fig. 2C). Based on this finding, POFUT2 has the potential to function as a carcinogenic gene in various types of cancer.

3.4 POFUT2 exhibits differential expression in pan-cancer cells and potentially predicts patient survival

OpenTarget was utilized to investigate the disease associated with POFUT2; the bubble graph illustrates that POFUT2 was linked to cancer and benign tumors (Fig. 3A). The disparities in POFUT2 mRNA levels between pan-cancerous and normal tissues were subsequently examined with TIMER2.0; POFUT2 mRNA expression was substantially increased in 10 distinct cancer types (GBM, LUAD, KIRC, LIHC, HNSC, LUSC, COAD, READ, CHOL, and STAD) (Fig. 3B). The disparities in POFUT2 mRNA levels between pan-cancerous and normal tissues were subsequently examined with GEPIA2.0; POFUT2 mRNA expression was substantially decreased in 9 distinct cancer types (CESC, LAML, LUSC, OV, READ, TGCT, THCA, and UCEC) and increased in GBM (Fig. 3C). For the prognostic significance of POFUT2, TCGA data were utilized to generate KM curves. We discovered that the POFUT2 upregulation was linked to lower OS percentages in ACC, BRCA, CESC, COAD, LGG, LIHC, and SARC; and lower DFS percentages of ACC, LIHC, and UVM (Fig. 3D). According to these findings, POFUT2 may serve as a cancer driver gene in ACC and LIHC and enhance the progression of BRCA, CESC, COAD, LGG, and SARC.

Fig. 3

POFUT2, associated with genomic instability, was differentially expressed, and predicted the survival of cancers. A The diseases associated with POFUT2 were analyzed on the openTarget web tool. The red dashed lines represent POFUT2-associated cancers. B The expression levels of POFUT2 mRNA in pan-cancer, and their corresponding control tissues were analyzed on TIMER2.0. Tumors and normal tissues are colored in red and blue, respectively, and SKCM metastasis tissues are in purple. C The expression levels of POFUT2 mRNA in pan-cancer, and their corresponding control tissues were analyzed on GEPIA2.0. Tumors and normal tissues are colored in red and green, respectively. D Prognosis prediction performance of POFUT2 in diverse cancers measured using GEPIA2. Survival heatmap for POFUT2 in 33 TCGA-derived cancers. Heatmap showing log10 HRs for POFUT2. Blue and red blocks represent decreased and increased risks, separately. Rectangles with frames indicate the signifcance upon prognosis analysis. OS overall survival, DFS disease-free survival

3.5 POFUT2 is involved in cancer immune infiltration and cytokine-mediated immune modulations

We analyzed the correlation between POFUT2 and 14 functional states in different tumors using CancerSEA data. The POFUT2 was positively correlated with angiogenesis, differentiation, hypoxia, inflammation, metastasis, and quiescence in OV, CRC, and RB et al. (P < 0.05) (Fig. 4A). However, POFUT2 was negatively associated with DNAdamage, DNArepair, EMT, invasion in OV, RB, UM et al. The cancers showing positively POFUT2 correlations with most immune checkpoint genes, including UVM, READ, DLBC, THYM, THCA, and SKCM et al. (Fig. 4B). However, POFUT2 was negatively associated with most immune checkpoint genes in TGCT. Finally, we compared the POFUT2 expression differences between pre- and post-cytokine treatment in cancer cell lines on web tool TISMO (Fig. 4C). We discovered that the POFUT2 expression decreased after the IFN-g treatment in four cell lines, and it also decreased in one IFN-b and one TNF-a posttreatment cell line. The results from multiple perspectives demonstrated that POFUT2 is a critical factor in immunosuppressive environment construction in many cancers, probably via suppressing immunostimulator function and immune checkpoint effects. Our analysis of POFUT2 expression in immunocytes allowed us to delve further into its role in cancer immunity. We first applied the CIBERSORT algorithm to obtain 22 immunocyte correlations with POFUT2. In several malignancies, we found that POFUT2 had a negative association with CD4 + memory resting T cells and resting NK cells, but a strong positive association with gamma delta T cells, CD8 + T cells, CD4 + memory active T cells, activated NK cells, and memory B cells. Furthermore, in six tumors, Tregs exhibited a positive association with POFUT2 (Fig. 4D).

Fig. 4

POFUT2 was reversely correlated to immune infiltration and cytokine interactions. A Correlation of the POFUT2 with 14 functional status in different tumors. Red represents positive correlation and blue represents negative correlation. B The heatmap of associations between immune checkpoints and POFUT2 expression in pan-cancer. C The multiple box plots of cancer cell lines POFUT2 expression pre- and post-cytokine treatment were retrieved from the TISMO web tool. D CIBERSORT calculation of immunocyte infiltration in pan-cancer. *, **, *** represents p < 0.05, p < 0.01, and p < 0.001, respectively

3.6 Mechanism of molecular docking-based drug binding to their targets

In light of the suboptimal therapeutic outcomes observed in patients with high-POFUT2 cancer who are undergoing standard chemotherapy, our objective was to identify prospective anti-POFUT2 drugs that could increase cancer sensitivity to existing chemotherapy while exerting greater effects. Compounds that induced transcriptional changes in nine distinct tumor cell lines that were contrary to those upregulated by high-POFUT2 expression were filtered with the cMap tool. The thirty compounds with the greatest potential to target POFUT2 were then presented (Fig. 5A). To determine whether these compounds are capable of binding to the POFUT2 protein, molecular docking was undertaken between the POFUT2 protein and potential drugs. The analysis encompassed the determination of the binding energy associated with every target-drug interplay by considering the binding modes exhibited by the targets and their potential drugs (Fig. 5). Strong hydrogen bonding and electrostatic interactions were the primary means by which each potential drug bound to its protein target, as shown by the results. Moreover, these potential drugs successfully occupied each target's active site. The binding energy for the POFUT2-Oligomycin-a complex is -9.5 kcal/mol, for the POFUT2-Rhapontin complex is -8.7 kcal/mol, for the POFUT2-Damnacanthal complex is -8.9 kcal/mol, and for the POFUT2-Antimycin-a complex is -9.1 kcal/mol, demonstrating exceptionally stable binding (Fig. 5B-5E). Taken together, Oligomycin-a, Rhapontin, Damnacanthal, and Antimycin-a were determined to be candidate POFUT2-targeted drugs and may be effective in cancer as alternative therapies.

Fig. 5

Binding mode of screened drugs to their targets by molecular docking. A The heatmap exhibits the top 30 compounds, experimentally causing transcriptional alterations opposite to those affected by median POFUT2 expression grouping. The color bar and the block color represent the similarity scores. B Binding mode of POFUT2-Oligomycin-a complex. C Binding mode of POFUT2-Rhapontin complex. D Binding mode of POFUT2-Damnacanthal complex. E Binding mode of POFUT2-Antimycin-a complex

3.7 Single-cell examination of the function of the FUT GSVA score

Ten functional states in various tumors were evaluated for their correlation with the FUT GSVA in our analysis. In ACC, the FUT GSVA scores were correlated positively with apoptosis, EMT, and hormone AR (p < 0.05, Fig. 6A).

Fig. 6

The FUT gene set variation analysis (GSVA) enrichment scores. A Correlation of FUT GSVA scores with 10 functional states in patients with different tumors. Red represents a positive correlation, and blue represents a negative correlation. *: P value ≤ 0.05; #: FDR ≤ 0.05. B Survival between the high and low GSVA score groups. C overall survival (OS) and D Progression-free survival (PFS) and E Disease-specific survival (DSS) of patients with ACC according to the GSVA score

To investigate the association of FUT expression with the survival of cancer patients, OS, PFS, DSS, and DFI were evaluated in patients stratified by FUT scores. A univariate analysis was implemented to ascertain the prognostic value of the FUTs (Fig. 6B). Patients exhibiting a high FUTs score had an unfavorable prognosis, as shown by the OS results for ACC (Fig. 6C) and MESO; this variable was determined to be a risk factor (all p < 0.05). According to the PFS findings, an increased FUTs score was linked to an unfavorable prognosis in ACC (Fig. 6D) and served as a patient risk factor (all p < 0.05). Additionally, an elevated FUT score was linked to an unfavorable DSS in ACC (Fig. 6E) and was determined to be a risk factor (all p < 0.05). Also, an elevated FUT score was linked to a dismal DFI in LUAD patients and was determined to be a risk factor (all p < 0.05). The results suggest that the FUTs score may potentially function as a conventional prognostic predictor and may be beneficial in anticipating outcomes for patients diagnosed with diverse tumor types, with particular emphasis on ACC.

3.8 Development of a risk prognostic model utilizing FUTs in ACC

By employing LASSO Cox regression analyses, FUTs with a positive impact on prognosis that were erroneously predicted were eliminated. As illustrated in Fig. 7A, B, the acquired FUTs comprised FUT11, FUT6, FUT4, FUT2, and FUT1. Five selected FUTs were utilized in the development of prognostic risk models via LASSO Cox regression analysis. FUT11, FUT6, FUT4, FUT2, and FUT1 were determined to be risk factors by this analysis. Risk scores were calculated as indicated below: risk score = (0.1203) * FUT1 + (0.3332) * FUT2 + (0.517) * FUT4 + (0.7169) * FUT6 + (0.1858) * FUT11.

Fig. 7

Construction of a prognostic risk model of FUTs for ACC patients. A The trajectory of each independent variable. The horizontal axis represents the log value of the independent variable lambda, and the vertical axis represents the independent variable’s coefficient. B Confidence intervals with different values of lambda. C Risk plot distribution, survival status of patients, and heatmap of the expression of the 5-gene signature in the whole TCGA-ACC dataset. D Kaplan–Meier survival curves for the risk model. (E) ROC curves for the risk model

After that, the median risk score was employed to divide patients into 2 categories: high-risk and low-risk. All patients with ACC were included in the distribution of survival status and risk plots, as shown in Fig. 7C; OS rates were considerably lower among high-risk patients in comparison with low-risk patients, according to the findings. Additionally, the gene heatmap demonstrated that the high-risk group had substantially greater levels of FUT11, FUT6, FUT4, FUT2, and FUT1 (Fig. 7C). Eventually, five FUTs, namely, FUT11, FUT6, FUT4, FUT2, and FUT1, were found to provide predictive value for OS. These 5 FUTs were employed to create a prognostic risk model. By applying the FUT signature to the entire dataset comprising 169 samples, its prognostic significance was validated. OS was found to be shortened in the high-risk patients relative to those at low risk, as determined by KM survival analyses (p = 0.000354; Fig. 7D). The AUC value for the FUT signature in the training cohort was 0.826, 0.813, 0.852, and 0.78 at 2, 4, 6, and 8 years (Fig. 7E). Overall, these outcomes show that the ACC prognostic risk model, which was developed utilizing the selected five FUTs, was accurate.

We determined whether differentially expressed FUTs exhibited a strong link to worse prognosis in ACC patients by conducting univariate analyses. This study set out to examine the entire TCGA dataset (n = 169) for any correlation between FUT mRNA expression levels and OS in ACC patients. Through the use of univariate analysis on the entire TCGA dataset (169 patients), we determined the predictive value of FUTs that showed differential expression in ACC patients. Elevations of the four differentially expressed FUTs (FUT11, FUT4, FUT2, and FUT1) were substantially linked to ACC patients’ OS (p < 0.05, Fig. 8A). Subsequently, it was determined that FUT4, FUT2, and FUT6 affected ACC patients' OS rates and clinical outcomes, as evidenced by multivariate analyses (p < 0.05, Fig. 8B). Moreover, OS was found to be strongly linked to pathological T stage, M stage, and TNM stage in ACC patients, according to univariate analysis (p < 0.05, Fig. 8A). Additionally, T stage and TNM stage were linked to ACC patients’s OS as evidenced by multivariate analysis (p < 0.05, Fig. 8B). The ACC patients’ prognoses were predicted using a nomogram (Fig. 8C) that integrated the risk score with the T stage. The C-index for survival prediction was 0.88 (P < 0.001). Furthermore, the nomogram model's calibration curves for 1, 2, 3, and 5 years exhibited a high degree of concordance between the nomogram's predicted values and the actual results (Fig. 8D). High accuracy in anticipating OS for patients with ACC was shown by the integrated prognostic nomogram that uses the FUT signature.

Fig. 8

Relationships between the risk model and overall survival rate, as well as clinicopathological features, among patients diagnosed with ACC. A Univariate and B multivariate Cox analyses of clinical characteristics. C The nomogram based on pathological T stage for predicting the prognosis of ACC patients. D Calibration plots for the overall survival nomogram model. The dashed diagonal indicates the ideal nomogram, and the purple line, the orange line, and the blue line represent the predicted 1-year, 2-year, 3-year, and 5-year overall survival, respectively, of ACC patients