Prediction of hepatocellular carcinoma prognosis and immunotherapeutic effects based on tryptophan metabolism-related genes

Abstract

Background: L-tryptophan (Trp) metabolism involved in mediating tumour development and immune suppression. However, comprehensive analysis of the role of the Trp metabolism pathway is still a challenge.

Methods: We downloaded Trp metabolism-related genes' expression data from different public databases, including TCGA, Gene Expression Omnibus (GEO) and Hepatocellular Carcinoma Database (HCCDB). And we identified two metabolic phenotypes using the ConsensusClusterPlus package. Univariate regression analysis and lasso Cox regression analysis were used to establish a risk model. CIBERSORT and Tracking of Indels by DEcomposition (TIDE) analyses were adopted to assess the infiltration abundance of immune cells and tumour immune escape.

Results: We identified two metabolic phenotypes, and patients in Cluster 2 (C2) had a better prognosis than those in Cluster 1 (C1). The distribution of clinical features between the metabolic phenotypes showed that patients in C1 tended to have higher T stage, stage, grade, and death probability than those of patients in C2. Additionally, we screened 739 differentially expressed genes (DEGs) between the C1 and C2. We generated a ten-gene risk model based on the DEGs, and the area under the curve (AUC) values of the risk model for predicting overall survival. Patients in the low-risk subgroup tended to have a significantly longer overall survival than that of those in the high-risk group. Moreover, univariate analysis indicated that the risk model was significantly correlated with overall survival. Multivariate analysis showed that the risk model remained an independent risk factor in hepatocellular carcinoma (p < 0.0001).

Conclusions: We identified two metabolic phenotypes based on genes of the Trp metabolism pathway, and we established a risk model that could be used for predicting prognosis and guiding immunotherapy in patients with hepatocellular carcinoma.

Keywords: HCC; Immune escape; Metabolic phenotype; Prognosis; Risk model; Trp metabolism.

Fig. 1

Identification of distinct metabolic phenotypes based on Trp-related genes. A The expression profile of Trp metabolism genes in TCGA-LIHC. B Forest plot of genes significantly correlated with prognosis. C The relative expression level of TPH1 measured by RT-qPCR in HCC cells and normal liver cell line. D Heatmap of correlation analysis of prognosis-related genes. E. CDF curve of samples. F CDF delta area curve of consensus clustering. G The sample clustering heatmap. H The survival analysis of the two subtypes in the TCGA-LIHC cohort and GSE14520 cohort. I Differences in Trp metabolism scores between C1 and C2 in the TCGA-LIHC cohort and GSE14520 cohort

Fig. 2

The differences in the immune cell infiltration characteristics and immunotherapy/chemotherapy response between C1 and C2. A Different infiltrating levels of 22 immune cells between the two molecular subtypes. B The differential expression of ICI genes between C1 and C2. C Differences in TIDE scores between C1 and C2

Fig. 3

Establishment of a novel risk model based on the DEGs between C1 and C2. A Volcano plot of DEGs. B The differentially expressed genes were analyzed by univariate regression. C The trajectory of each independent variable with lambda. D Confidence interval under lambda. ROC curve and survival analysis were used to construct a risk model in the TCGA-LIHC dataset (E) in the GSE76427 dataset (F)

Fig. 4

Immune cell infiltration characteristics in distinct risk subgroups. A Boxplot of differences in the infiltrating abundance of 22 immune cells between different risk subgroups. B Boxplots of differences in immune scores calculated between the risk subgroups by the ESTIMATE method. C Correlation between 22 immune cell components and risk score. D Heatmap of enrichment scores of pathways. E Correlation analysis between risk score and the pathways (R > 0.7). F The correlation between the risk score and the tryptophan metabolism pathway

Fig. 5

The risk model has excellent predictive power for immunotherapy and chemotherapy for HCC. A Differentially expressed immune checkpoint genes between different risk subgroups. B Differences in TIDE scores between different risk subgroups. C Correlation between risk score and TIDE scores

Fig. 6

Clinical application of risk models for predicting prognosis and response to immunotherapeutic effect. A Univariate Cox analysis of clinical characteristics and RiskType based on TCGA database. B Multivariate Cox analysis of clinical characteristics and RiskType. C In the IMvigor210 cohort, SD/PD patients had higher risk scores than other types of responders. D The percentage statistics showed that the treatment effect was significantly better in the low-risk group than in the high-risk group. E Prognostic difference in risk subgroups in the whole TCGA-LIHC cohort. Prognostic difference in early-stage patients in the IMvigor210 cohort. Prognostic difference between different risk groups of early-stage patients (F) and late-stage patients in the IMvigor210 cohort (G)