Comprehensive analysis of transcriptome and metabolome analysis in Intrahepatic Cholangiocarcinoma and Hepatocellular Carcinoma

Abstract

Intrahepatic cholangiocarcinoma (ICC) and hepatocellular carcinoma (HCC) are liver originated malignant tumors. Of the two, ICC has the worse prognosis because it has no reliable diagnostic markers and its carcinogenic mechanism is not fully understood. The aim of this study was to integrate metabolomics and transcriptomics datasets to identify variances if any in the carcinogenic mechanism of ICC and HCC. Ten ICC and 6 HCC who were resected surgically, were enrolled. miRNA and mRNA expression analysis were performed by microarray on ICC and HCC and their corresponding non-tumor tissues (ICC_NT and HCC_NT). Compound analysis was performed using capillary electrophoresis time-of-flight mass spectrometry (CE-TOFMS). Principle component analysis (PCA) revealed that among the four sample groups (ICC, ICC_NT, HCC, and HCC_NT) there were 14 compounds, 62 mRNAs and 17 miRNAs with two distinct patterns: tumor and non-tumor, and ICC and non-ICC. We accurately (84.38%) distinguished ICC by the distinct pattern of its compounds. Pathway analysis using transcriptome and metabolome showed that several pathways varied between tumor and non-tumor samples. Based on the results of the PCA, we believe that ICC and HCC have different carcinogenic mechanism therefore knowing the specific profile of genes and compounds can be useful in diagnosing ICC.

Figure 1. Hierarchical Clustering using correlation coefficients.

Hierarchical clustering of 96 PCs: consisting of 32 PCs each obtained from mRNAs, miRNAs and compounds. Each PC consists of 32 dimensional vectors with 32 elements, each of which corresponds to the contribution of each sample to each PC. The correlation coefficients between PCs were computed using these 32 elements. On the Vertical axis are absolute negative correlation coefficients that are used as distance for hierarchical clustering (lower pairs have larger absolute correlations). Red rectangle indicates 5PCs which were chosen by hierarchical clustering.

Figure 2. Scatter diagram.

Lower triangle: Scatter plot between five PCs (a) PC3_comp, (b) PC1_mRNA, (c) PC2_mRNA, (d) PC1_miRNA, and (e) PC2_miRNA, selected based on the hierarchical clustering shown in Fig. 1. Each row and column corresponds to the PC displayed diagonally. Open rectangle in the second row and third column indicates the correlation coefficients (upper numerical value) and their P-values (lower numerical value) associated with the corresponding scatter plots.

Figure 3. Box plot in 5 PCA.

Box Plot of PCs is shown in Fig. 2. The sample contribution to each PC are shown using boxplot classified based on four sample classes. The correlation coefficients used for the hierarchical clustering tree represented in Fig. 1 and shown in the upper triangle in Fig. 2 are for the distances found among the five plots. Vertical line represents genetic distance. Red dots are depicted as measured value of each probe after normalization. P values are shown in a scatter plot (P value of PC3_comp, PC1_mRNA, PC2_mRNA, PC1_miRNA, and PC2_miRNA, correspond to 8.68e-03, 1.69e-02, 3.98e-06, 4.74e-02, and 9.42e-03, respectively).

Figure 4. Embedding of compound, mRNA, and miRNA.

(a) Two-dimensional embedding of compounds spanned by PC2 and PC3. (b) Two-dimensional embedding of mRNA spanned by PC1 and PC2. (c) Two dimensional embedding of miRNA spanned by PC1 and PC2. We chose compounds, mRNA and miRNAs as “outliers” for further analysis.

Figure 5. Box plot of selected compounds.

Amount of 14 compounds used to discriminate between tumor (ICC and HCC) and non-tumor (ICC_NT and HCC_NT) or between ICC and non-ICC (ICC-NT, HCC, and HCC-NT). Left vertical axis shows the amount of compound. Red dots are depicted as measured value of each probe after normalization. Each p-value is indicated in Table S2. XC00061 is only known as Pubchem accession number (http://www.ncbi.nlm.nih.gov/pccompound). Unknown: neither the name of component nor the Pubchem accession number is known.

Figure 6. Box plot of selected mRNAs.

Expression pattern of 62 mRNA used to discriminate between tumor (ICC and HCC) and non-tumor (ICC_NT and HCC_NT) or between ICC and non-ICC (ICC-NT, HCC, and HCC-NT). In cases where there are multiple probes for one mRNA, pleural box plots with the same mRNA are listed. Vertical left axis shows the expression level of mRNAs. Red dots are depicted as measured value of each probe after normalization. P-values are indicated in Table S3.

Figure 7. Box plot of selected miRNAs.

Expression pattern of 17 miRNAs used to discriminate between tumor (ICC and HCC) and non-tumor (ICC_NT and HCC_NT) or between ICC and non-ICC (ICC-NT, HCC, and HCC-NT). Vertical left axis shows miRNA expression levels. P-values are indicated in Table S4.