3.1 The Cox analysis to investigate the prognostic significance of AS events in HCC cohort

In a comprehensive study involving 343 patients with HCC, a total of 26,157 AS events originating from 12,817 genes were identified. On average, each gene exhibited two AS events. Among the various types of AS events, ES was the most prevalent, followed by AT and AP events. Specifically, 9409 ES events were detected across 4070 genes, 6009 AT events were observed in 2636 genes, and 4874 AP events were identified in 1971 genes. Additionally, 2141 AA events were found in 1538 genes, 1865 AD events were present in 1308 genes, 1795 RI events were detected in 1231 genes, and 64 ME events were observed in 63 genes (Fig. 1).

Fig. 1

Schematic representation of AS patterns and AS events in HCC. A Diagram illustrating different AS events, including AA, AD, AP, AT, ES, RI, and ME. B Count of AS events and associated genes in HCC

To evaluate the prognostic significance of AS events in HCC patients, Cox univariate survival analysis was performed. This analysis aimed to assess the impact of each AS event on OS. The results indicated that out of the 26,157 AS events identified, 3164 of them were found to be significantly associated with OS in HCC patients (p < 0.05). These significant AS events may potentially serve as prognostic markers for HCC patients. For more detailed information on the specific AS events and their associations with OS, please refer to Table s01.

Figure 2 depicts the 20 most notable AS events associated with survival across the seven types of AS events. A clear observation from the figure is that the majority of these AS events are linked to unfavorable prognostic outcomes. Notably, it is evident that a single gene in HCC can exhibit multiple AS events that are associated with survival. To visually represent the overlap between the seven types of AS events in HCC, researchers generated an UpSet plot (Fig. 3A). Remarkably, a significant proportion of genes associated with survival exhibited at least two types of AS events, with some genes even demonstrating four types of AS events. For instance, the CCL14 gene displayed significant associations between its AA, AP, ES, and RI events and OS. In order to investigate the functional relationships among these crucial survival-associated events (p < 0.005), a gene interaction network was constructed using the STRING Database (Fig. 3B). This network analysis provides valuable insights into potential molecular interactions and functional connections among the genes involved in these AS events, thereby shedding light on the underlying mechanisms associated with the prognosis of HCC.

Fig. 2

Forest plots generated to display the subgroup analyses of survival-associated AS events in the HCC cohort. A–G depicted the hazard ratios (HRs) for the top 20 survival-associated events in different AS types, including AA, AD, AP, AT, ES, RI, and ME events in HCC. The color scale of the circles represented the corresponding p-values, while the horizontal bars represented the 95% confidence intervals (CIs)

Fig. 3

The analysis of survival-associated AS events in the HCC cohort. A UpSet intersection diagram displaying seven types of survival-associated AS events in the HCC cohort. B Gene network illustrating the interactions among survival-associated AS events in HCC, created using Cytoscape

These findings provide valuable insights into the prognostic value of AS events in HCC patients and offer clues for further research.

3.2 Cox regression model analysis of prognostic factors in HCC cohort

To uncover autonomous prognostic determinants within individuals diagnosed with HCC, we cautiously handpicked the most momentous AS phenomena associated with survival as potential contenders. We methodically assembled multivariable Cox regression models for each of the seven distinct variants of candidate AS events, aimed at identifying autonomous prognostic indicators. By ingeniously amalgamating the diverse array of candidate AS events from these seven categories, we fabricated the ultimate prognostic predictor. Notably, our meticulous examination of the data unequivocally demonstrated the potent prognostic capacities of all seven prognostic models, each predicated on disparate AS event types, in predicting the prognosis of HCC patients (Fig. 4A–G).

Fig. 4

Kaplan–Meier plot and ROC curves to assess the survival rate for seven types of AS events as prognostic predictors in the HCC cohort. A–G represents a specific type of AS event, with the high-risk group depicted in red and the low-risk group depicted in blue. H A Kaplan–Meier plot to visualize the survival probability over time for the final prognostic predictor in the HCC cohort. The high-risk group is shown in red, while the low-risk group is shown in blue. I ROC analysis to evaluate the performance of all prognostic predictors in the HCC cohort. The ROC curves of the prognostic predictors are color-coded to represent different types of AS events

Remarkably, it is worth noting that among all the diverse AS events examined (Fig. 4A–G), the prognostic model centered around a solitary AD unveiled the most exceptional performance. Building upon this noteworthy finding, we proceeded to fabricate the ultimate prognostic predictor by amalgamating disparate AS events from all seven distinct types—an endeavor that yielded unprecedented outcomes. It is of paramount significance to emphasize that the performance of this final prognostic predictor surpassed that of each individual splicing pattern, as clearly illustrated in Fig. 4H. Notably, in the analysis of the ROC, the final prognostic predictor exhibited an impressively high AUC of 0.830. Additionally, the AD model and the AA model closely trailed behind, with AUCs of 0.819 and 0.816, respectively, as portrayed in Fig. 4I. These findings serve as compelling evidence that the amalgamation of the various prognostic models within the final combination considerably enhances the predictive accuracy concerning the prognosis of HCC patients. It is also worth mentioning that Table 1 provides a comprehensive inventory of the 15 HCC-specific genes corresponding to the AS events that are encompassed within the final amalgamation of prognostic models.

Table 1 HCC-specific genes corresponding to the AS events in the prognostic model3.3 Interaction analysis between prognostic-related AS events and splicing factors

In order to determine the splice events associated with survival in HCC patients, we performed survival analysis of splice factors based on gene expression levels. The results showed that 50 splice factors were significantly correlated with OS. Additionally, using Spearman’s test, we examined the correlation between the PSI values of the most important AS events and the expression of survival-associated splice factors (Fig. 5A). We found that most adverse survival-associated AS events (red dots) were positively correlated with splice factor expression (gray dots) (red line), while the most favorable prognostic AS events (orange dots) were negatively correlated with splice factor expression (purple line). The high expression of splice factor PRPF38B was associated with poorer patient survival, whereas the high expression of splice factor NONO was associated with better prognosis (Fig. 5B, C). The scatter plot displayed the correlation between splice factor PRPF38B and the AS of RTN4 (Fig. 5D), indicating a positive correlation between high PRPF38B expression and poorer OS. Similarly, the scatter plot showed the correlation between splice factor NONO and the AS of CLSPN (Fig. 5E), suggesting a negative correlation between high NONO expression and favorable prognosis.

Fig. 5

A correlation network to visualize the relationship between splicing factors and genes in the HCC cohort. A The correlation network between the expression levels of survival-associated splicing factors and the Percent Spliced In (PSI) values of AS genes in the HCC cohort. B, C Kaplan–Meier curves generated to assess the survival probability over time for the splicing factors PRPF38B and NONO, respectively. The high-expression group is represented by the red curve, while the low-expression group is represented by the blue curve in the HCC cohort. D The correlation between the expression level of the splicing factor PRPF38B and the PSI value of the alternative splicing gene RTN4. E The correlation between the expression level of the splicing factor NONO and the PSI value of the alternative splicing gene CLSPN

3.4 Identification of FTCD as a significant gene for HCC by a comprehensive pan-cancer analysis

Through meticulous analysis of Table s01 and extensive review of relevant literature, we have successfully identified the FTCD gene as a compelling candidate that exhibits downregulation in HCC cases and demonstrates a positive correlation with favorable patient prognosis. Subsequently, we delved deeper into our investigation by scrutinizing the expression patterns of FTCD across multiple cancer types, including Endometrial Cancer, Cervical Cancer, Miscellaneous Neuroepithelial Tumor, Sarcoma, and Non-Small Cell Lung Cancer. Our findings, as depicted in Fig. 6, illustrate the distinctive and specific expression of FTCD in HCC.

Fig. 6

The pan-cancer analysis of FTCD demonstrates its specific expression in the liver. A Displays the TPM (Transcripts Per Million) statistics for the gene. B, C provide statistics on the types of mutations. D Illustrates the mutation sites of FTCD in pan-cancer

Consequently, we leveraged comprehensive data including copy number variation (CNV), RNA-Seq data, methylation status, and clinical information from the HCC cohort in the TCGA database to construct a survival curve for FTCD. Our analysis revealed that patients exhibiting high FTCD CNVs, elevated FTCD expression, and reduced methylation levels experienced more favorable prognoses (as depicted in Fig. 7A). This compelling evidence supports the notion that FTCD holds promise as a specific prognostic factor for HCC.

Fig. 7

HCC multi-omics prognostic features. A Kaplan–Meier plot illustrating the survival probability over time based on the prognostic predictors of CNV, FPKM and methylation of FTCD. The high-expression group is represented by the red curve, while the low-expression group is represented by the blue curve. B, C Present statistics on the types of mutations. D Displays the mutation sites. E Shows the distribution of mutation types in terms of expression levels. F Represents the distribution of expression levels for methylation sites. G Illustrates the distribution of expression levels for methylated sites. H Depicts the distribution of AS expression levels

To gain further insights, we meticulously examined the mutation types and positions within the FTCD gene. Our investigations indicated that the majority of mutations present were missense mutations, with the most prevalent mutation types being C > T and G > A (as shown in Fig. 7B). Delving into the mutation type distribution of FTCD in HCC, we expanded our analysis to include liver cancer data from various databases such as IRKEN, MSK, AMC, and INSERM (Fig. 7C). Strikingly, our findings unveiled the presence of specific structural variations (SV) in the FTCD gene solely in HCC cases, while no such SV occurrences were observed in other cancer types (as portrayed in Fig. 6C). This strongly implies a direct association between FTCD SV and the development of HCC.

Subsequently, utilizing the Linked Omics of Lots of Indels and Polymorphisms in Cancer (LOLLIPOP) tool, we investigated fusion genes involving FTCD. Notably, we identified the specific expression of the E323Sfs*55 fusion in HCC cases, whereas the E251* mutation demonstrated higher prevalence across pan cancers (depicted in Figs. 6D and 7D, respectively). These observations suggest that SV events generate fusion genes, thereby altering the functional characteristics of the FTCD gene with specific relevance to HCC.

We scrutinized the expression patterns of SV and mRNA z-scores (normalized to diploid samples) for the FTCD gene, unveiling elevated expression levels in cases exhibiting increased copy numbers and the presence of SV (as illustrated in Fig. 7E). Additionally, the distribution plot of methylation sites (Fig. 7F) illustrated varying degrees of methylation across different exons. Specifically, low levels of methylation were observed in the initial exons, while the later exons demonstrated higher levels. Notably, distinct differences in methylation were observed among different tumor stages. In stage II and IIIa, aberrantly elevated methylation levels were detected in the initial exons of some patients. Moreover, our investigation into the distribution of FTCD gene methylation across different stages of HCC revealed variations in expression levels and methylation extent (as demonstrated in Fig. 7G). Analyzing AS events of the FTCD gene, we identified significant differences in expression patterns of the fifth exon among different subtypes. This finding strongly reinforces the association between FTCD's AS patterns and HCC (as depicted in Fig. 7H).We conducted KEGG enrichment analysis and protein–protein interaction network analysis for the FTCD gene. The enrichment analysis revealed an enrichment of the PPAR pathway, which is associated with tumors. The protein–protein interaction network analysis resulted in three modules that were clustered using K-means algorithm (Fig. 8).

Fig. 8

KEGG enrichment analysis and protein–protein interaction (PPI) network analysis of co-expressed genes with FTCD. A Enrichment analysis reveals the enrichment of the PPAR pathway, which is associated with tumors. B PPI analysis using K-means clustering results in the formation of three modules

3.5 Validation the role of FTCD in the biological characteristics of liver cancer cells and its molecular mechanisms in vivo and in vitro

In our cell model experiments, a notable decrease in both gene expression and protein levels of FTCD was observed in the HCC group compared to the Control group (p < 0.05) (as depicted in Fig. 9A–C). Through CCK-8 experiments, we additionally discovered that the HCC group exhibited the highest cell proliferation capacity, while the Control group displayed the weakest (Fig. 9D). Remarkably, the FTCD-overexpressing group (FTCD group) demonstrated a significantly attenuated cell proliferation ability in comparison to the HCC group, albeit still higher than that of the Control group (p < 0.05). This finding strongly suggests the inhibitory role of the FTCD gene in regulating cell proliferation. Furthermore, our investigation delved into the expression levels of genes associated with the PPAR/PI3K/AKT/mTOR pathway, as shown in Fig. 8. Notably, by overexpressing FTCD (FTCD group) in BEL-7402 cells, we documented a significant downregulation in the expression of PPARδ, PIK3CA, and AKT, along with an elevation in the expression of the upstream negative regulator PTEN (p < 0.05) (Fig. 9A–C). These compelling findings reinforce the hypothesis that FTCD may exert its inhibitory effect on the biological characteristics of liver cancer cells through the precise regulation of the PPAR/PI3K/AKT/mTOR pathway.

Fig. 9

Validation the role of FTCD in cell model and mice model. A qPCR results of FTCD, PPARδ, PIK3CA, AKT, and PTEN genes in the cell model. B, C Western blot results of FTCD, PPARδ, PIK3CA, AKT, and PTEN in the cell model. D Cell proliferation results using the CCK-8 assay. E Immunohistochemistry results in the mice model. F qPCR results of FTCD, PPARδ, PIK3CA, AKT, and PTEN genes in the mice model. G, H Western blot results of FTCD, PPARδ, PIK3CA, AKT, and PTEN in the mice model

In the mice model, researchers introduced liver cancer cells transfected with FTCD overexpression plasmids into mice to establish a tumor-bearing mice model. The growth rate and size of the tumors were carefully observed and compared among different groups, while liver tissues were collected for pathological examination. Immunohistochemistry was employed to evaluate the staining of the FTCD protein in the mice liver tissues. Significantly, the content of FTCD protein in the liver tissues of the tumor-bearing mice model (H group) was found to be lower than that in the CK group (p < 0.01). However, in the FTCD overexpression tumor-bearing mice model (F group), the FTCD protein content in the liver tissues was higher than that in the H group (p < 0.05) (as demonstrated in Fig. 9E). Furthermore, using RT-PCR detection, we observed that the expression level of FTCD in the control group (CK group) surpassed that in both the tumor-bearing mice model (H group) and the FTCD overexpression tumor-bearing mice model (F group) (p < 0.05) (Fig. 9F). Consistently, western blot analysis revealed that the content of FTCD protein in the control group (CK group) was higher compared to both the H group and F group (p < 0.05) (Fig. 9G, H), corroborating the results obtained from the pathological examination. Importantly, these findings highlight the crucial role of FTCD in liver cancer and its correlation with molecular mechanisms.

In summary, our study encompassed several key steps to provide a comprehensive understanding of the significance of FTCD in liver cancer and its associated molecular mechanisms. Initially, we performed a genome-wide analysis using RNA-seq data from TCGA to investigate prognostic-associated AS events within a cohort of 343 liver cancer patients. Through comprehensive Cox proportional hazards regression analysis, we established a prognostic prediction model. By examining the correlation between survival-associated AS events and splicing factors, we constructed a splicing network to elucidate the underlying regulatory mechanisms. Subsequently, we identified a notable association between FTCD and prognosis, prompting further exploration of AS patterns within the FTCD gene. To validate the inhibitory effect of FTCD on the biological characteristics of liver cancer cells, we employed both mice models and cell models. The findings from these experiments supported the notion that FTCD exerts its inhibitory influence by modulating the PPAR/PI3K/AKT/mTOR pathway.

In essence, this comprehensive investigation sheds light on the pivotal role of FTCD in liver cancer and provides insights into the intricate molecular mechanisms underlying its function.