Histology of liver injury and HCC
To follow the development of liver injury, liver tissue from Mdr2 −/− mice were studied at 12, 21 and 42 weeks after birth (Fig. 1a). To control for effect of ageing, WT mice were studied at 12 weeks (baseline/WT) and 42 weeks (aged/WT) (Fig. 1a). H&E, Reticulin and Masson’s trichrome stained liver sections confirmed normal liver at baseline/WT, portal inflammation at 12 weeks, cirrhosis at 21 weeks and advanced cirrhosis with presence of HCC at 42 weeks (Fig. 1b). As expected, 100% of Mdr2 −/− mice had macroscopically visible tumours at 42 weeks (Fig. 1b) with a mean number of 17.5 (± 3.9) visible tumours that measured 3 to 19 mm in diameter. Normal liver was seen at 42 weeks in WT mice (aged/WT) (histology not shown).
Microbiome clustering occurs with stage of disease
The fecal microbiome of all Mdr2 −/− and WT mice was profiled by 16S rRNA gene amplicon sequencing. Illumina sequencing produced a total of 5,460,928 sequences, with an average of 116,062 sequences per sample, post-quality filtering.
Clustering of microbiome communities occurred with stages of liver injury and HCC in Mdr2 −/− mice. Significant separation of microbial communities occurred between all stages of progressive liver injury and with HCC development as assessed by Bray–Curtis dissimilarity matrix, shown in Principal Coordinate Analysis (PCoA) plot (P < 0.001) (Fig. 2a). The baseline/WT and aged/WT microbiome were clustered together and were distinctly separate from Mdr2 −/− mice, thus confirming that separation of microbial communities in Mdr2 −/− mice was a result of liver injury and HCC development rather than ageing effect (Supporting Fig. 1a, Additional File 1).
Alpha diversity as calculated by Shannon’s index was significantly reduced in liver cirrhosis (P = 0.038) and HCC (P = 0.022) when compared to baseline/WT mice (Fig. 2b). Additionally, Shannon’s index was reduced in liver cirrhosis (P = 0.004) and HCC (P = 0.002) compared to the inflammation timepoint; however, there was no difference between cirrhosis and HCC (Fig. 2b). Other indices of alpha diversity (including observed OTUs) also changed across the spectrum of liver injury, but no consistent pattern emerged (Supporting Fig. 2, Additional File 1). Importantly, no change in alpha-diversity was seen in WT mice with ageing (baseline/WT vs aged/WT mice), confirming that reduced alpha-diversity observed in Mdr2−/− mice was a result of liver injury and HCC development rather than ageing effect (Supporting Fig. 2, Additional File 1).
Microbial composition shifts with progressive liver injury and HCC development
Sequences were classified into six phyla, accounting for 99.8% of total phyla level abundance (Fig. 2c). A clear difference in community structures at the phylum level was seen based on stage of liver injury and HCC development. Four of six measured phyla were significantly different between the different stages. Firmicutes were measured at a lower abundance during inflammation (P = 0.015), cirrhosis (P < 0.0001) and HCC (P = 0.0004), when compared to the baseline/WT (Fig. 2c and Supporting Fig. 3, Additional File 1). The composition of the microbiome in the inflammation stage was not significantly different from the cirrhosis stage, apart from an increase in relative abundance of Bacteroidetes during the cirrhosis time point (P = 0.004) (Fig. 2c and Supporting Fig. 3, Additional File 1). The HCC disease stage was characterised by an increase in the relative abundance of Tenericutes (P = 0.009) and Actinobacteria (P = 0.004) compared to the cirrhosis stage (Fig. 2c and Supporting Fig. 3, Additional File 1). No difference between phyla was observed in WT mice with ageing (baseline/WT vs aged/WT mice) (Supporting Fig. 1b, Additional File 1).
At genus level, key taxa were enriched at the various stages of injury and HCC. During liver inflammation, there was an increase in Staphylococcus compared to cirrhosis (P = 0.005) and HCC time points (P = 0.002) (Fig. 3a). Also, an increase in Pediococcus was seen in inflammation compared to all other time points (all P < 0.0001) (Fig. 3b). In liver cirrhosis, an increase in Prevotella was seen compared to baseline/WT (P = 0.0001) and HCC (P = 0.0003) (Fig. 3c). Additionally, in liver cirrhosis, Bacteroides was enriched compared to baseline/WT (P = 0.024) and HCC (P = 0.028) (Fig. 3d). Importantly, Parabacteroides became more enriched with progressive liver injury/disease, with peak relative abundance occurring in HCC compared to baseline/WT (P < 0.001) and inflammation (P = 0.016) but not cirrhosis time points (Fig. 3e). Clostridium emerged in HCC time point, being significantly enriched compared to all other time points (all P < 0.05) (Fig. 3f). Detailed data is shown in Supporting Table 1, Additional File 2.
Microbial functional capacity shifts with progressive liver injury and HCC development
A range of functional pathways pertaining to the microbiome were found to be significantly increased in each of the phases of liver injury/disease in Mdr2 −/− mice. The time point of inflammation and cirrhosis were enriched with microbiome pathways related to gut barrier dysfunction, namely bacterial invasion of epithelial cells and glycosaminoglycan degradation compared to baseline/WT (all P < 0.05) (Fig. 4a-b and Supporting Fig. 4, Additional File 1). Functional pathways related to lipopolysaccharide (LPS) biosynthesis were elevated in cirrhosis (P = 0.001) and HCC (P = 0.003) compared to baseline/WT (Fig. 4d and Supporting Fig. 4, Additional File 1).
With respect to metabolic pathways, a theme reflecting the rearrangement of the cellular energy source was evident. Compared to baseline/WT, during liver inflammation and cirrhosis, there was an increase in microbial functional pathways related to metabolism of carbohydrates (glycolysis and starch/sucrose metabolism), which then significantly declined from cirrhosis to HCC (all P < 0.0001) (Fig. 4e-f and Supporting Fig. 4, Additional File 1). In contrast, functional pathways related to amino acid degradation (valine, leucine, isoleucine and phenylalanine metabolism) increased at the HCC time point compared to all other time points (all P < 0.0001) (Fig. 4g-h and Supporting Fig. 4, Additional File 1). Detailed data is shown in Supporting Table 2, Additional File 2.
Progressive liver injury and HCC development is associated with increased serum LPS and a shift toward a Th1/Th17 proinflammatory cytokine milieu
In the serum, LPS levels increased with progressive liver injury and peaked with development of HCC in Mdr2 −/− mice. Serum LPS levels were higher at the HCC time point compared to both inflammation and cirrhosis timepoints (P < 0.0001 and P = 0.002, respectively) (Fig. 5a). In parallel, a shift toward a Th1/Th17 proinflammatory cytokine milieu was seen with HCC development. The HCC timepoint was characterized by a significant increase in interferon-gamma (IFN-γ), tumour necrosis factor alpha (TNF-α), IL-6 and IL-17 compared to inflammation and cirrhosis timepoints (all P < 0.05) (Fig. 5b-e). Other proinflammatory cytokines followed a similar pattern (Supporting Fig. 5, Additional File 1). The converse was seen with the anti-inflammatory cytokine, IL-10, which was significantly reduced in the serum at the HCC timepoint compared to inflammation and cirrhosis timepoints (P = 0.022 and P = 0.032, respectively) (Fig. 5f). Detailed data is shown in Supporting Table 3, Additional File 2.
Transcriptional profiling of mouse liver tissue demonstrates alterations in intrahepatic inflammatory responses across the spectrum of liver injury to HCC
Evaluation of the expression of genes throughout the spectrum of liver injury in Mdr2 −/− mice (n = 4 per group) showed a significant increase in intrahepatic inflammatory responses during inflammation and cirrhosis time points, which was followed by immunosuppression with the development of HCC (Supporting Fig. 6, Additional File 1). Upregulation of type I (Ifn) encoding genes (Ifna2 and Ifnar1), important in first line defense against microbial invasion and loss of immune tolerance [10] were seen in inflammation and cirrhosis stages compared to all other timepoints (all P < 0.01) (Fig. 6a). Genes activated in response to microbial components Tlr3 and Tlr4 followed a similar pattern, elevated during inflammation and cirrhosis stages compared to the HCC timepoint (all P < 0.01) in Mdr2 −/− livers (Fig. 6b). In parallel, genes activated by LPS including myeloid differentiation factor 88 (Myd88) and interferon regulatory factor 3 (Irf3) were upregulated in inflammation and cirrhosis compared to advanced cirrhosis and HCC (all P < 0.001) (Fig. 6c). Several proinflammatory mediators, for example TNF-α and IL-1β (Fig. 6d) and CCR5, followed the same pattern in Mdr2 −/− livers (all P < 0.01).
In contrast to the predominant proinflammatory milieu detected at the inflammation and cirrhosis time points, a blunted inflammatory response prevailed in advanced cirrhosis and HCC. When comparing advanced cirrhosis with HCC time points, a further reduction in several proinflammatory cytokines including Tlr4 (Fig. 6b) and IL-1β (Fig. 6d) was seen (P = 0.0037 and P = 0.0032, respectively). There was, however, ongoing overexpression of Tlr1, Tlr2 (Fig. 6e) and Tlr6 genes throughout all time points including HCC compared to baseline/WT. Tlr1, Tlr2 and Tlr6, are known to recognize lipopeptides and to mediate cross-desensitization to LPS [11, 12] and their expression coincided with the emergence of microbial functional pathways relating to bacterial invasion of the gut barrier and increased LPS biosynthesis pathways in addition to increased serum LPS levels.
Furthermore, genes typically expressed by monocytes/macrophages, CD14 and Irf7, were overexpressed throughout the course of liver injury with further elevation of CD14 in advanced cirrhosis compared to both inflammation and cirrhosis time points (P = 0.037 and P = 0.041, respectively) (Fig. 6f). Detailed data is shown in Supporting Table 4, Additional File 2.
Changes in intrahepatic inflammatory responses correlate with changes in gut microbial composition and serum LPS level
To identify relationships between gut microbiome composition and the intrahepatic immune response, correlation analysis was performed between microbial abundance (genus level) and expression of intrahepatic genes across all stages of liver injury and HCC in Mdr2 −/− mice. Additionally, given its integral importance to both microbial and intrahepatic inflammatory responses, serum LPS levels were included in our correlation analysis.
Following correction for multiple testing (FDR > 0.25), Spearman’s correlation analysis revealed a number of host intrahepatic genes that covaried with gut microbial abundance in Mdr2 −/− mice (Fig. 7). Overall, changes in the expression levels of key genes had both positive and negative correlations with changes in abundance of various genera primarily from the phyla Bacteroidetes, Firmicutes and Proteobacteria. To this effect, Prevotella (phylum Bacteroidetes) enriched in cirrhosis was positively correlated with Tlr2, a gene elevated in cirrhotic Mdr2−/− livers and important in LPS cross-desensitization (R = 0.65, P = 0.041) (Fig. 7). Enrichment of Parabacteroides (phylum Actinobacteria) seen at the HCC time point was negatively correlated with the inflammatory gene Ifnb1 (R = − 0.65, P = 0.038) (Fig. 7).
Serum LPS levels, shown to increase throughout the time course of liver disease (Fig. 5a), were found to be negatively correlated with Tlr4 (R = − 0.97, P = 0.033) and Myd88 (R = − 0.71, P = 0.041) (Fig. 7); genes which are well established to be involved in macrophage mediated LPS immunotolerance.