Identification of a novel FOXO3‑associated prognostic model in hepatocellular carcinoma

Abstract

Although numerous molecular classifications are available to predict the prognosis of patients with hepatocellular carcinoma (HCC), they are still unsatisfactory. Forkhead box O3 (FOXO3) has been widely reported as a transcription factor involved in human cancers, but its role in HCC remains controversial. The present study aimed to explore the role of FOXO3 in HCC, as well as to identify biomarkers and construct prognostic models based on FOXO3. FOXO3 was highly expressed in HCC and was closely associated with poor prognosis in The Cancer Genome Atlas (the training set) and International Cancer Genome Consortium (the validation set). Subsequently, a co-expression network indicated that the red modules were closely related to FOXO3. Five key FOXO3-related genes [DEAD-box helicase 55 (DDX55), RAB10, member RAS oncogene family (RAB10), RAB7A, TATA-box binding protein associated factor, RNA polymerase I subunit B (TAF1B) and TAF3] were obtained using Cox-least absolute shrinkage and selection operator analyses. The 5-gene signature successfully predicted the prognosis of patients with HCC in both the training and validation sets. Enrichment analysis suggested marked differences in AKT and cell cycle-related (E2F targets and G₂/M checkpoints) pathways between HCC subgroups. Furthermore, the tumor microenvironment analysis suggested that the difference in the distribution of M2 macrophages among various subgroups may contribute to the poor prognosis using the CIBERSORTx framework. Furthermore, the mRNA and protein expressions of DDX55, RAB10, RAB7A, TAF1B and TAF3 were found to be higher in HCC tissues compared with paracancerous tissues using RT-qPCR and western blotting. Additionally, knockdown of RAB10, RAB7A and TAF3 inhibited proliferation of Huh7 cells, assessed by a Cell Counting Kit-8 assay. In conclusion, a novel FOXO3-related model was constructed and revealed that RAB10, RAB7A and TAF3 may be potential molecular targets or biomarkers for HCC.

Keywords: Forkhead box O3; bioinformatics; biomarkers; hepatocellular carcinoma; prognosis.

Figure 1.

FOXO3 is highly expressed in HCC tissues and associated with poor prognosis. (A) mRNA expression of FOXO3 in HCC and normal tissues from TCGA dataset. (B) mRNA expression of FOXO3 in pathological grading G1/2 and G3/4 from TCGA dataset. (C) mRNA expression of FOXO3 in HCC and normal tissues from the ICGC dataset. (D) mRNA expression of FOXO3 in HCC with or without bile duct invasion from the ICGC dataset. Overall survival in patients with HCC with high or low FOXO3 expression from (E) TCGA and (F) ICGC datasets. FOXO3, Forkhead box O3; HCC, hepatocellular carcinoma; TCGA, The Cancer Genome Atlas; ICGC, International Cancer Genome Consortium; BIHCC, bile duct invasion hepatocellular carcinoma.

Figure 2.

Weighted correlation network analysis of The Cancer Genome Atlas dataset. (A) Analysis of the scale-free index (left) and mean connectivity (right) for various soft-threshold powers (β). (B) Dendrogram of all differentially expressed genes clustered based on the measurement of dissimilarity (1-TOM). The color band shows the distribution of modules. (C) Heatmap of the correlation between the modules and mRNA expression of FOXO3 in HCC. The module with the highest coefficient was used for subsequent analysis. (D) Module membership vs. gene significance scatter plot of FOXO3. FOXO3, Forkhead box O3; HCC, hepatocellular carcinoma.

Figure 3.

Identification of the 5-gene signature. (A) Forest plot of univariate Cox regression analysis in The Cancer Genome Atlas dataset. Cox results for the top five and bottom five genes. (B) Least absolute shrinkage and selection operator regression analysis. (C) λ curves show the least absolute shrinkage and the best λ was selected based on the minimum criteria. (D) The protein-protein interaction network between FOXO3 and 294 genes. FOXO3, DDX55, RAB10, RAB7A, TAF1B and TAF3 are highlighted by red dots and the remaining genes are represented by green dots. The color of the edges was determined by the combined score obtained from STRING. FOXO3, Forkhead box O3; DDX55, DEAD-box helicase 55; RAB10, RAB10, member RAS oncogene family; RAB7A, RAB7A, member RAS oncogene family; TAF1B, TATA-box binding protein associated factor, RNA polymerase I subunit B; TAF3, TATA-box binding protein associated factor 3; SLC4A1AP, solute carrier family 4 member 1 adaptor protein; ZNF765, zinc finger protein 765; MORC3, MORC family CW-type zinc finger 3; WWC2, WW and C2 domain containing 2; UBE3A, ubiquitin protein ligase E3A; SECISBP2L, SECIS binding protein 2 like; FAN1, FANCD2 And FANCI associated nuclease 1.

Figure 4.

Construction and validation of a 5-gene prognostic model. (A) The distribution of RS, survival status and expression of the 5-gene signature between the high- and low-risk groups in TCGA dataset. (B) Overall survival of patients with HCC in the high- and low- RS groups in TCGA dataset. (C) Time-dependent ROC curve of RS in TCGA dataset. (D) The distribution of RS, survival status and expression of the 5-gene signature between the high- and low-risk groups in the ICGC dataset. (E) Overall survival of patients with HCC in the high- and low-RS groups in the ICGC dataset. (F) Time-dependent ROC curve of RS in the ICGC dataset. RS, risk score; TCGA, The Cancer Genome Atlas; ICGC, International Cancer Genome Consortium; ROC, receiver operating characteristic; DDX55, DEAD-box helicase 55; RAB10, RAB10, member RAS oncogene family; RAB7A, RAB7A, member RAS oncogene family; TAF1B, TATA-box binding protein associated factor, RNA polymerase I subunit B; TAF3, TATA-box binding protein associated factor 3; AUC, area under the curve.

Figure 5.

Clinical potential of the RS for prognosis. (A) Univariate and (B) multivariate Cox analyses of the RS and other clinical factors in TCGA dataset. (C) Nomogram of TNM staging and RS based on the Cox results in TCGA dataset. (D) Univariate and (E) multivariate Cox analyses of the RS and other clinical factors in the ICGC dataset. (F) Nomogram of TNM staging and RS based on the Cox results in the ICGC dataset. RS, risk score; TCGA, The Cancer Genome Atlas; ICGC, International Cancer Genome Consortium; TNM, tumor, node, metastasis.

Figure 6.

Analyses of the carcinogenic pathways and tumor microenvironment between the high- and low-RS groups in The Cancer Genome Atlas dataset. Gene set enrichment analysis of the high-RS group vs. the low-RS group for (A) the G₂/M checkpoint, (B) E2F targets and (C) PI3K/AKT/MTOR signaling. (D) Distribution of 22 immune cell types in the tumor microenvironment of each hepatocellular carcinoma sample. (E) Comparison of 22 immune cell types infiltrated in tumor microenvironment between the high- and low-RS groups. -, not significant; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001. RS, risk score; FDR, false discovery rate; NES, normalized enrichment score.

Figure 7.

Identification of five biomarkers in HCC. (A) mRNA expression of DDX55, RAB10, RAB7A, TAF1B and TAF3 in HCC and normal tissues from TCGA. Overall survival curves in patients with HCC with high or low expression of (B) DDX55, (C) RAB10, (D) RAB7A, (E) TAF1B and (F) TAF3 from TCGA dataset. The best cut-off value for each gene was obtained using the X-tile software. (G) mRNA expression of DDX55, RAB10, RAB7A, TAF1B and TAF3 in HCC and paired paracancerous tissues from 10 patients with HCC, assessed using reverse transcription-quantitative PCR. (H) Immunohistochemical images of DDX55, RAB10 and RAB7A in HCC and normal tissues obtained from the HPA. (I) Protein expression of DDX55, RAB10, RAB7A, TAF1B and TAF3 in HCC and paired paracancerous tissues from 10 patients with HCC, evaluated using western blot analysis. The figure on the right is a relative semi-quantitative histogram of protein expression. The gray values were obtained by ImageJ and analyzed using GraphPad Prism. (J) Cell Counting Kit-8 curves after knockdown of RAB10, RAB7A and TAF3 in Huh7 cells by siRNA. *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001. T, tumor; NT, not-tumor; HCC, hepatocellular carcinoma; TCGA, The Cancer Genome Atlas; DDX55, DEAD-box helicase 55; RAB10, RAB10, member RAS oncogene family; RAB7A, RAB7A, member RAS oncogene family; TAF1B, TATA-box binding protein associated factor, RNA polymerase I subunit B; TAF3, TATA-box binding protein associated factor 3; HPA, The Human Protein Atlas; NC, negative control.