Construction and validation of a novel IGFBP3-related signature to predict prognosis and therapeutic decision making for Hepatocellular Carcinoma

Abstract

Background: IGFBP3 plays a pivotal role in carcinogenesis by being anomalously expressed in some malignancies. However, the clinical value of IGFBP3 and the role of IGFBP3-related signature in HCC remain unclear.

Methods: Multiple bioinformatics methods were used to determine the expression and diagnostic values of IGFBP3. The expression level of IGFBP3 was validated by RT-qPCR and IHC. A IGFBP3-related risk score (IGRS) was built via correlation analysis and LASSO Cox regression analysis. Further analyses, including functional enrichment, immune status of risk groups were analyzed, and the role of IGRS in guiding clinical treatment was also evaluated.

Results: IGFBP3 expression was significantly downregulated in HCC. IGFBP3 expression correlated with multiple clinicopathological characteristics and demonstrated a powerful diagnostic capability for HCC. In addition, a novel IGRS signature was developed in TCGA, which exhibited good performance for prognosis prediction and its role was further validated in GSE14520. In TCGA and GSE14520, Cox analysis also confirmed that the IGRS could serve as an independent prognostic factor for HCC. Moreover, a nomogram with good accuracy for predicting the survival of HCC was further formulated. Additionally, enrichment analysis showed that the high-IGRS group was enriched in cancer-related pathways and immune-related pathways. Additionally, patients with high IGRS exhibited an immunosuppressive phenotype. Therefore, patients with low IGRS scores may benefit from immunotherapy.

Conclusions: IGFBP3 can act as a new diagnostic factor for HCC. IGRS signature represents a valuable predictive tool in the prognosis prediction and therapeutic decision making for Hepatocellular Carcinoma.

Keywords: Biomarker; Diagnostic; Hepatocellular carcinoma; IGFBP3; Prognostic model.

Figure 1. The flow diagram of this study.

Figure 2. Expression of IGFBP3 in hepatocellular carcinoma.

(A) IGFBP3 mRNA levels between LIHC and normal tissues in TCGA. (B) Expression status of IGFBP3 in GTEx normal, TCGA normal, and TCGA-LIHC tissues. (C–E) Relative expression of IGFBP3 in LIHC tissues and in normal tissues in the GSE54236 (C), GSE14520 (D), and GSE76427 (E) datasets. (F) RT-qPCR showed decreased IGFBP3 mRNA levels in HCC cell line (Huh7). (G) The protein expression of IGFBP3 in LIHC specimens and normal liver specimens from CPTAC datasets. (H) Typical images of IHC showing the protein expression of IGFBP3 in HCC and adjacent non-tumor tissues. (**p < 0.01, ***p < 0.001).

Figure 3. Diagnostic value of IGFBP3 and its relevance to clinical features.

Boxplots demonstrating the expression of IGFBP3 in patients are grouped according to clinical characteristics. (A) Age; (B) Gender; (C) T stage; (D) N stage; (E) M stage; (F) pathologic stage; (G) histologic grade; (H) vascular invasion (^nsp > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001). (I–L) ROC curve of IGFBP3 in LIHC based on TCGA-LIHC (I), GSE14520 (J), GSE76427 (K), and ICGC-LIRI (L).

Figure 4. Construction of IGFBP3-related risk score (IGRS).

(A) Venn diagram indicating the 10 IGFBP3-related genes identified in three cohorts. (B) Construction of the LASSO model based on IGFBP3 and its related genes. (C) The optimal λ of the LASSO model. (D) The risk factor diagram of IGRS model in TCGA cohort. (E) The OS curve for high- and low- IGRS groups in TCGA cohort. (F) 1-, 3-, and 5-year ROC curves of IGRS model for survival prediction in TCGA cohort. (G) The distribution and median cutoff value of IGRS, the OS status of each sample, and the expression value of the eight model genes in the GSE14520 dataset. (H) The prognostic significance of IGRS in GSE14520 cohorts. (I) Time-dependent ROC analyses of the IGRS regarding the OS and survival status in the GSE14520 cohort.

Figure 5. Univariate and multivariate Cox regression analysis of prognosis in HCC patients.

The univariate and multivariate Cox regression analyses in (A) TCGA cohort and in (B) ICGC cohort.

Figure 6. Nomogram to evaluate the OS probability based on TCGA cohort.

(A) The nomogram for predicting the 1-, 3- and 5-year OS probabilities. (B) Comparison of C-index among age, gender, grade, stage, LRRS, and nomogram. (C–E) Calibration curves of the nomogram to predict (C) 1-, (D) 3- and (E) 5-year OS probabilities. (F–H) Decision curve analysis (DCA) among the age, gender, grade, stage, LRRS, and nomogram with respect to the (F) 1-, (G) 3-, and (H) 5-year OS.

Figure 7. Functional enrichment analyses between the high- and low-IGRS groups.

(A) Volcanic map of DEGs between the high and low IGRS groups. (B) Heat map for top 60 DEGs between high and low IGRS subgroups. (C–D) The results of GSEA (KEGG pathways) in the high-IGRS (C) and low-IGRS groups(D).

Figure 8. Immune profile and prediction of treatment style of IGRS-based HCC groups.

(A) The significant immune-associated pathways in the high-IGRS group. (B–F) Differences in ImmuneScore (B), estimated scores (C), eight common immune checkpoint genes (D), chemokines and receptors (E), and MHC molecules (F) between the two risk groups, respectively. (G) The landscape of immune cell infiltration between two IGRS subtypes estimated by the ssGESA. ns ≥ 0.05, * < 0.05, ** < 0.01, and *** < 0.001.