The proportion of serum secondary bile acids in HCC patients was decreased
The levels of serum total BAs in the HCC group (5251±1460 nM) were higher than those in the healthy controls (4626±1015 nM), but this difference did not reach statistical significance (P>0.05) (Figure 1a). Surprisingly, there was a significantly lower ratio of secondary BAs to primary BAs in the serum of the HCC group than in the healthy control group (0.18±0.02 vs 0.52±0.11, P = 0.008) (Figure 1b). Specifically, the percentages of conjugated and unconjugated secondary BAs in the total BAs were both significantly reduced in the HCC group (11.0%±1.3% vs 17.1%±2.4%, P = 0.028, and 3.8%±0.7% vs 12.4%±2.8%, P=0.009) (Figure 1c and Figure 1d).
The proportion of secondary bile acids was reduced in DEN-HCC mice
To further verify the hypothesis that the percentage of secondary BAs in the total BAs was decreased in HCC, we generated mouse liver cancer models through the combinative induction of diethylnitrosamine (DEN) and hepatotoxin carbon tetrachloride (CCl4) (Figure 2a). We found that the level of serum total BAs was significantly increased in the DEN-HCC mouse group compared to that in the control mouse group (11244±3690 nM vs 1556±407 nM, P <0.001, Figure 2b). Notably, the ratio of secondary BAs to primary BAs was also remarkably reduced in the DEN-HCC mouse group (0.24±0.04 vs 0.60±0.14, P = 0.029, Figure 2c). Similarly, compared with those of the normal control mouse group, the percentages of conjugated and unconjugated secondary BAs of the DEN-HCC mouse group were also decreased (3.6%±0.7% vs 10.3%±1.1%, P<0.001, and 14.9%±2.5% vs 24.1%±3.5%, P = 0.048, Figure 2d and 2e).
Characterization of gut microbiome compositional profiles in HCC patients
It is well known that the gut microbiota can affect the metabolism of bile acids and change the composition of bile acids [8]. To display microbiome β-diversity, we used principal coordinate analysis (PCoA) coupled with unweighted UniFrac distances and found a clear separation between HCC patients and healthy controls (Figure 3a). Moreover, to display the overlaps between two groups, we used a Venn diagram and observed that 1262 of the 2109 OTUs were shared between the two groups (Figure 3b). Notably, 699 of 1961 OTUs were unique to HCC patients, while only 148 of 1410 OTUs were unique to healthy persons (Figure 3b).
Among the bacterial compositions, the bacterial phyla Bacteroidetes, Firmicutes, Proteobacteria, and Actinobacteria were the most abundant bacteria. Compared with healthy controls, Actinobacteria was significantly decreased in HCC patients (P =0.02). In addition, Bacteroidales, Lactobacillales, Selenomonadales, Verrucomicrobiales, and Enterobacteriales were increased in HCC, while Clostridiales, Fusobacteriales, Pasteurellales, and Burkholderiales were decreased in HCC, but these differences between them were not statistically significant (Figure 3c). At the order level, probiotic Bifidobacteriales, belonging to the phylum Actinobacteria, was significantly decreased in HCC patients (P=0.026, Figure 3d). The fecal microbial composition in each sample from the two groups at the phylum and order levels is presented in Figures S1C and S1D. Since the production of secondary bile acids requires the participation of BSH enzymes from Bifidobacteriales, Lactobacillales, Bacteroidales, and Clostridiales [50], we checked the abundance of BSH-rich bacteria in HCC patients (76.6%±4.0%) and observed that it was lower than that in healthy controls (80.3%±2.5%); however, their difference did not reach statistical significance (P=0.462, Figure 3e and 3f).
Characterization of gut microbiome compositional profiles in mice with HCC
Likewise, to further verify the HCC-related changes in the gut microbiome, we collected fecal samples from mice with DEN-induced HCC and controls. After using PCoA to display microbiome β-diversity, we found two distinct enterotypes between the two groups (Figure 4a). Furthermore, a Venn diagram showed that 437 of the 609 OTUs were shared between the two groups. Notably, 109 of 546 OTUs were unique to mice with HCC, while only 63 of 500 OTUs were unique to control mice (Figure 4b).
In addition, the bacterial phyla Bacteroidetes, Firmicutes, Proteobacteria, and Actinobacteria were still the most abundant bacteria in the two groups (Figure 4c). Compared with control mice, phylum unidentified Bacteria was significantly increased in HCC (P=0.004), and Proteobacteria was significantly decreased in HCC (P<0.001, Figure 4c). At the order level, Erysipelotrichales (P=0.002), unidentified Bacteria (P=0.006), and Coriobacteriales, (P=0.02), were remarkably increased in mice with HCC, while Clostridiales (P=0.005), Desulfovibrionales (P<0.001), and Enterobacteriales (P=0.007) were significantly decreased in HCC (Figure 4d). The fecal microbial composition in each sample from the two groups at the phylum and order levels is presented in Figure S2C and S2D. We also found that the abundance of BSH-rich bacteria in DEN-induced HCC mice (70.7%±6.5%) was markedly lower than that in normal control mice (91.0%±0.6%, P =0.007)(Figure 4e and Figure 4f). Therefore, we hypothesized that the decreased abundance of BSH-rich bacteria may be involved in the production of secondary bile acids, which is related to the development of HCC.
Antibiotic vancomycin decreased the abundance of BSH-rich bacteria, lowered the levels of secondary BAs, and induced tumor growth
To further confirm our hypothesis that the decrease in BSH-rich bacteria is involved in the development of HCC through downregulating the levels of secondary BAs, we used vancomycin to treat C57BL/6 mice and then generated orthotopic implanted liver tumor models. We found that the tumor weight in the vancomycin treatment group was higher than that in the control group (P =0.075, Figure 5a). Furthermore, we used 16S rDNA to analyze the gut microbiota between the two groups and found that the abundance of BSH-rich bacteria in the vancomycin treatment group was significantly lower than that in the control group (20.0%±3.4% vs 93.0%±2.2%, P=0.009, Figure 5b and 5c). To further observe the effect of vancomycin treatment on serum bile acids, we found that the concentration of serum total BAs in the vancomycin treatment group was higher than that in the control group (3895±1495 nM vs 3026±1079 nM), but the difference was not statistically significant (P=0.644, Figure 5d). Interestingly, the ratio of secondary BAs to primary BAs of vancomycin-treated mice was significantly lower than that of the control group (0.05±0.01 vs 0.35±0.11, P=0.032, Figure 5e). However, the percentage of conjugated secondary BAs in vancomycin-treated mice (5.1%±1.0%) was not significantly lower than that in control mice (6.3±1.5%), though the percentage of unconjugated secondary BAs was significantly reduced in vancomycin-treated mice compared with control mice (0.1%±0.1% vs 17.8%±6.5%, P=0.02, Figure 5f and Figure5g).
Conjugated and unconjugated secondary bile acids are both reduced
The above results indicated that the reduction in the secondary BA rate was closely associated with the development of HCC. It is well known that hydrophobic TCA, LCA, DCA and CDCA are cytotoxic, while hydrophilic BAs are cytoprotective, such as UDCA and TUDCA [18]. Therefore, we focused on hydrophilic secondary BAs and conjugated secondary BAs. In humans, secondary BAs are mainly composed of GUDCA, UDCA, GDCA and DCA. In mice, secondary BAs include TUDCA, UDCA, TDCA, DCA, THDCA and HDCA. We found that the percentages of GDCA (P=0.003) and DCA (P =0.007) in HCC patients were decreased significantly (Figure 6a-6d). Similarly, in the DEN-induced HCC mice, the percentages of TUDCA, TDCA, DCA, and THDCA were all significantly reduced, and the P values were <0.001, 0.003, 0.022, and <0.001, respectively (Figure 7a-7f). Further analysis found that the percentages of UDCA, TDCA, DCA, THDCA and HDCA in the vancomycin-treated mice were significantly reduced, and the P values were 0.033, 0.015, 0.030, 0.024, and 0.029, respectively (Figure 8a-8f). Through the above data, we observed a remarkable reduction of serum conjugated DCA (a kind of conjugated secondary BAs) in HCC patients, DEN-induced HCC mice and vancomycin-treated mice, further deduced that conjugated DCA might be intimately associated with HCC progression.
GDCA inhibits the growth and migration of HCC cells
As a kind of common conjugated DCA in humans, GDCA was used to treat human HCC cell lines, including SUN-449 and HepG2 cell lines. We found that GDCA markedly decreased the clone formation rates of SUN-449 and HepG2 cell lines (P<0.05 and P<0.01) compared with LO2 human hepatocyte cell lines (Figure 9a). Using CCK-8 assays, we similarly demonstrated that the proliferation of SUN-449 cells and HepG2 cells was significantly inhibited 3 d, 4 d, and 5 d after GDCA treatment. The inhibitory rates of SUN-449 cells were 51.4%, 42.7%, and 44.1% (P<0.05, P<0.05, and P<0.01), respectively. The inhibitory rates of HepG2 cells reached 50.9% (P<0.05), 66.3% (P<0.05), and 64.4% (P<0.01). However, the growth of LO2 cells was not inhibited by GDCA (Figure 9b). Next, we studied the effect of GDCA on the migration of SUN-449 cells and HepG2 cells by Transwell assays and wound healing assays and found that GDCA significantly blocked the ability of SUN-449 cells and HepG2 cells to migrate through the membrane and refill an empty area (“scratch”, Figure 9c and 9d). In addition, we used Annexin V tests to examine the effect of GDCA on the apoptosis of HCC cells (SUN-449 cells and HepG2 cells) and observed that after GDCA treatment, the apoptosis rates of SUN-449 cells and HepG2 cells were remarkably increased (12.4%±0.46% vs 28.8%±0.28%, in SUN-449 cells, P<0.05 and 5.6%±0.37% vs 22.1%±6.32%, in HepG2 cells, P<0.05, Figure 9e).