The Intestinal Microbiome, Plasma Metabolome and Liver Transcriptome- A Conspiracy Driving Hepatic Steatosis

With chronic overnutrition the storage of triglyceride in numerous cells including hepatocytes seems inevitable, and can be explained by well recognized pathways of lipid metabolism. This seemingly obvious association between chronic overnutrition and hepatocyte steatosis appears to be regulated by an unnervingly complex set of pathways that are not traditionally thought of as metabolic regulators. The study of regulation of steatosis is traditionally approached by hypothesis driven manner which examine the role of an individual molecules and pathways. Such studies however cannot provide an overview of multiple pathways, identify which pathways are proximal, and which are dominant. To gain such information large data sets on multiple pathways are required, and such data sets can be analyzed in a variety of ways. In data driven analysis there is no pre-analysis hypothesis, and large data sets are used to identify pathways which are associated with the phenotype of interest. A major limitation of this approach is that choices in the analysis methodology can identify very different pathways as being important. This makes it vital that results from data driven studies are validated in additional patient populations, and by direct experimentation.

The recent manuscript by Hoyles et al reports a data driven study which obtained a comprehensive data set on a population with non-alcoholic fatty liver disease (NAFLD)(1). This data includes stool metagenomics, plasma and urine metabolomics and whole liver transcriptomics. The conclusions from this data driven analysis are supplemented by human to mouse fecal microbiota transplant (FMT), and in vitro testing of individual metabolites on primary human hepatocytes. Before delving into what was found it is important to highlight that the population is not representative of NAFLD, and sample collection is not representative of the steady state. The study population was limited to women (Spanish n=44, and Italian n=61) who had had undergone preparation for bariatric surgery. In addition to the lack of men, the process for selection for bariatric surgery may introduce relevant biases such as enrichment with patients with a greater concern for obesity related co-morbidities. Further, stool and body fluids were obtained weeks apart, with liver tissue obtained at the time of surgery and other samples a few weeks before. The months leading up to bariatric surgery involve significant medical weight loss intervention with most patients losing between 5%–10% of total body weight. This results in a reduction in liver volume and steatosis, and relevant to this paper changes plasma metabolome and stool microbiome(2). These issues makes it unlikely that samples in the study are representative of NAFLD patients in steady state. There were also a few unexpected omissions such as mean BMI at the time of sample collection, degree of pre-operative weight loss, and stage of liver fibrosis.

Initial analysis was between pairs of data sets from the large data set of steatosis, intestinal microbiome, plasma metabolome and liver transcriptome, followed by analysis if the identified associations were stable over multiple data sets. There was a robust reverse correlation between a low microbial gene richness (MGR) as defined by the average number of gene counts from 7 million DNA reads, and hepatic steatosis, alanine aminotransferase and C-reactive protein, and this was independent of the BMI. In the plasma metabolome, there was a strong positive correlation between the plasma levels of branched-chain amino acids (leucine, valine, isoleucine), and a weak negative correlation of plasma choline and phosphocholine with steatosis, which extended to a correlation with low MGR. In the liver transcriptome approximately 3000 genes were positively and 3000 negatively correlated with steatosis. 1,800 genes were positively correlated with high MGR with a significant overlap between this set and the 3000 genes that were negatively correlated with steatosis. A gene pathway analysis, with a focus on the 1,800 genes that were positively associated with high MGR and negatively associated with steatosis highlighted a variety of immune pathways as well as pathways associated with the proteasome, phagosome, insulin resistance, glucagon signaling and responses to bacterial and viral infections.

To test if the microbiota can be causal in the development of steatosis, and if this is via changes in the plasma metabolome an FMT experiment was performed in mice. Stool from three donors with grade 3 steatosis and three donors with grade 0 steatosis was gavaged into wild-type mice who had received 7 days of antibiotic treatment, and were kept on regular diet for 2 weeks after FMT. There was a significant increase in total liver triglyceride content, and a statistical association between donor microbiome content and mouse plasma branch chain amino acids and trimethylamine N-oxide. Finally based on the entire data set phenylacetic acid (a catabolite of phenylalanine) was identified for potential pro-steatotic effects. Addition of phenylacetic acid to primary human hepatocyte cultures resulted in greater steatosis and the upregulation of lipid metabolizing genes. The production of phenylacetic acid from phenylalanine can occur by enzymes present in humans, and may occur by bacterial enzymes.

Collectively this is a very impressive data set, albeit in a highly pre-selected population. The cross analysis of the data sets has identified reduced microbiome diversity, certain liver transcriptomic pathways and branch chain amino acids as a set of closely integrated responses that are present in patients with NAFLD. The two sets of experiments to test the predictions from the analysis are less convincing. The FMT experiments result in minimal change in hepatic steatosis. It is difficult to assess if this can be of biological significance, but certainly serves as a proof-of-concept of something that is possible under experimental condition. The experiments on the ability of phenylacetic acid to increase hepatocyte steatosis lack an important control of another metabolite. It is not clear if the increase steatosis in hepatocytes is specific to phenylacetic acid, or would have occurred even if a metabolite predicted not to induce steatosis had been added to hepatocytes. The authors propose a hypothesis which places the intestinal microbiome as a key accelerator of liver steatosis by modifying the metabolome into one with pro-steatotic effects. This is an interesting proposal, and adds to the established concept of the microbiome providing pro-inflammatory products (liposaccharide being the primary one). Is this a valid overarching hypothesis to explain high levels of hepatic steatosis, or an overreaching set of conclusions? In support of the authors hypothesis previous studies have reported that the microbiome in NAFLD is different from obesity without NAFLD, and among patients with biopsy-proven NAFLD, the microbiome is different in those who have advanced fibrosis versus those with stage 0–2 fibrosis, demonstrating a unique disease-microbiome association (3–5). Moreover, recent data from a twin-study provides proof-of-concept data implicating the role of gut-microbial metabolites (such as 3–4 hydroxy phenyllactate) in linking shared gene effects between liver fat and fibrosis in NAFLD(6). Also, the total microbial genome is significantly greater than human genome and the function of most bacterial genes is not known (7–9). This certainly provides the microbiome with the potential to generate metabolic intermediates which may have pathological effects, and changes in the metabolome after manipulation of the microbiome are recognized(2). Detracting from the hypothesis is the fact that analysis of large data sets inevitably results in some positive associations, and the finding in this case are certainly novel and will need to be robustly verified experimentally as well as in diverse and independent cohorts of patients with biopsy-proven NAFLD. Furthermore, the key clinical question remains to be determined whether gut microbiome is causal in progression of hepatic steatosis to NASH to cirrhosis and hepatocellular carcinoma, and if either the microbiome or the microbiome derived metabolites may have a therapeutic potential to reverse disease progression in human NAFLD.

Figure. Gut microbiome and NAFLD.

The manuscript by Hoyles et al demonstrates associations that are consistent across the data sets B, C, D and E. It further shows that transfer of the intestinal microbiome from humans with steatosis to mice can result in a mild increase in liver triglycerides, and that some plasma metabolites can result in steatosis of hepatocytes in culture.