Integrated microbiome and metabolome analysis reveals the interaction between intestinal flora and serum metabolites as potential biomarkers in hepatocellular carcinoma patients

Abstract

Globally, liver cancer poses a serious threat to human health and quality of life. Despite numerous studies on the microbial composition of the gut in hepatocellular carcinoma (HCC), little is known about the interactions of the gut microbiota and metabolites and their role in HCC. This study examined the composition of the gut microbiota and serum metabolic profiles in 68 patients with HCC, 33 patients with liver cirrhosis (LC), and 34 healthy individuals (NC) using a combination of metagenome sequencing and liquid chromatography-mass spectrometry (LC-MS). The composition of the serum metabolites and the structure of the intestinal microbiota were found to be significantly altered in HCC patients compared to non-HCC patients. LEfSe and metabolic pathway enrichment analysis were used to identify two key species (Odoribacter splanchnicus and Ruminococcus bicirculans) and five key metabolites (ouabain, taurochenodeoxycholic acid, glycochenodeoxycholate, theophylline, and xanthine) associated with HCC, which then were combined to create panels for HCC diagnosis. The study discovered that the diagnostic performance of the metabolome was superior to that of the microbiome, and a panel comprised of key species and key metabolites outperformed alpha-fetoprotein (AFP) in terms of diagnostic value. Spearman's rank correlation test was used to determine the relationship between the intestinal flora and serum metabolites and their impact on hepatocarcinogenesis and progression. A random forest model was used to assess the diagnostic performance of the different histologies alone and in combination. In summary, this study describes the characteristics of HCC patients' intestinal flora and serum metabolism, demonstrates that HCC is caused by the interaction of intestinal flora and serum metabolites, and suggests that two key species and five key metabolites may be potential markers for the diagnosis of HCC.

Keywords: biomarkers; hepatocellular carcinoma; integrated analysis; intestinal flora; serum metabolites.

Figure 1

A schematic of the design and the experimental flow diagram. After a strict pathological diagnosis and exclusion process, 68 patients with HCC, 33 patients with cirrhosis, and 34 healthy controls were included at the Second Hospital of Nanjing, Jiangsu Province, China. In total, 132 serum samples and 134 feces samples were included in the analysis. Characterized the gut microbiome of 67 patients with HCC, 33 patients with cirrhosis, and 34 healthy controls and identified the microbial markers. Simultaneously characterized the serum metabolites from 66 hepatocellular carcinomas, 32 cases of hepatic sclerosis, and 34 healthy controls to identify metabolite markers. Random forest analysis is used to assess the ability of various marker combinations to distinguish the HCC cohort from the non-HCC cohort (cirrhosis and healthy controls). Using serum- and fecal-matched cohorts to examined the link between the gut microbiota and serum metabolites that changed significantly in HCC. HCC, hepatocellular carcinoma; KGM, key gut microbes; KSM, key serum metabolites; KGMSM, key gut microbial-associated serum metabolites.

Figure 2

The gut microbiome community is divided into three groups. (A) Rarefaction curves between the number of samples and the number of species. In all samples, the number of species approached saturation. Fecal microbial alpha diversity at the species level was estimated by the Chao1 index (B), Shannon index (C), and Simpson index (D). * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, ns p > 0.05. (E, F) The top 5 representative phyla and genera, as well as their proportions in each of the three groups. (G) The top 10 representative species and their proportions among the three groups. (H) A Venn diagram displaying group overlaps revealed that 274 of the total richness of 6640 species were unique to HCC. The red circle represents HCC, the blue circle represents LC, and the green circle represents NC.

Figure 3

Key species selection by LEfSe. Differential microbial score chart: the higher the score, the greater the contribution of the microbe to the difference.

Figure 4

Overview of altered serum metabolism in HCC (n=66) and non-HCC (LC (n=32), NC (n=34)). (A, B) PLS-DA shows the differences between the groups’ metabolites. The abscissa (PC1) and the ordinate (PC2) are the two main coordinates that explain the greatest difference between the samples. The number is the score of the principal component, which represents the percentage of the explanation on the overall variance of the specific principal component. The graph points represent samples, and different colors represent various sample grouping information; similar samples are clustered together. (C, D) The two rightmost points in the figure are the actual R2Y and Q2 values of the model, and the remaining points are the R2Y and Q2 values obtained by randomly arranging the samples used. This result is mainly used to judge whether the model is overfitting and the validity of the model. A volcano plot is a graphical representation of differential metabolism. (E) Metabolites that differ between HCC and NC. (F) Metabolites that differ between HCC and LC. Green marks the downregulated differential metabolites, red marks the upregulated differential metabolites, and metabolites without differences are labeled purple−gray.

Figure 5

(A–C) Bubble plots of pathways with significant enrichment of differential metabolites. The ordinate is the name of the metabolic pathway, and the abscissa is the rich factor (rich factor = the number of differential metabolites annotated to the pathway/all identified metabolites annotated to the pathway). The larger the rich factor, the greater the proportion of differential metabolites annotated to the pathway. The color from blue to red indicates that the p value decreases sequentially; the larger the point, the more differential metabolites are enriched in the pathway. (D) Distribution of different metabolites in each group. *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 6

(A, B) The heatmap of the top 20 differential species and differential metabolites with the smallest p values for every omics in HCC vs. non-HCC. Columns represent the differential metabolites, and rows represent the differential species. The color blocks represent the correlation coefficient. The darker the color, the stronger the correlation between the different species and the different metabolites. Red represents a positive correlation, and blue represents a negative correlation. * represents p<0.05, ** represents p<0.01.

Figure 7

Correlation analysis of the metagenome and metabolome. (A) Spearman correlation network interaction diagram of the key species and differential metabolites. Each point in the figure represents a species or a metabolite. The more lines there are between the points, the more species or metabolites it may regulate. Blue dots represent species, red dots represent metabolites, red connecting lines between dots are positive correlations, and green lines are negative correlations. The thickness of the line represents the level of the correlation coefficient. (B, C) Random forest ROC map of species and metabolomes (the ROC map of the metabolome is on the left, the ROC map of species is in the middle, and the ROC map of the species and metabolomes is on the right).

Figure 8

(A–F) The ROC curves of a random forest analysis of different panels. KGM, key gut microbes; KSM, key serum metabolites; KGMSM, key gut microbial-associated serum metabolites. The abscissa of the ROC curve is the false-positive rate; the ordinate is the true positive rate; the blue curve is the average curve after 10 folds; the AUC is the area under the curve; the shaded region is the upper and lower 1 standard deviation.

Figure 9

(A) Spearman correlation chord diagram of key species and key metabolites. Species or metabolites are on the edge of the circle in the figure, and the connecting line in the circle represents the correlation between the species and metabolites; red is a positive correlation, and blue is a negative correlation. The darker the color or the thicker the line, the stronger the correlation. (B, C) Pearson correlation cluster heatmap depicting the relationships between the key metabolites (ouabain, TCDCA, GCDCA, theophylline, and xanthine), the key species (Odoribacter splanchnicus and Ruminococcus bicirculans) and the clinical indicators. WBC, white blood cell; LY, leukocytes; GGT, gamma-glutamyl transferase; EOS, eosinophil; TP, total protein; PLT, platelets; NE, neutrophilic granulocyte; TBIL, total bilirubin; MONO, monocytes; BASO, basophil. Red indicates positive correlations, whereas blue indicates negative correlations. *p < 0.05, **p < 0.01, ***p < 0.001.