As NAFLD is a broad-spectrum disease with pathologies ranging from simple steatosis to cirrhosis or HCC 17, it is difficult to distinguish between the stages of NAFLD without liver biopsy, which is an invasive procedure. miRNAs are novel candidates for potential diagnostic and prognostic biomarkers of NAFLD because they are present in abundance in the liver and are associated with various kinds of liver diseases.18. We observed that several miRNAs showed significant differences inexpression in accordance with the severity of NAFLD and histological findings.
In NAFLD patients diagnosed with NASH or advanced fibrosis, the fibrosis stage is determined to be 3 or 4, and the prognosis is typically poor compared to that of simple steatosis.3. In patients with NAFLD, NASH represents a severe form of hepatic damage due to the recruitment of pro-inflammatory immune cells, and it can eventually progress to cirrhosis and hepatocellular carcinoma.19 Steatosis evaluation is important for the diagnosis of NAFLD, and change in steatosis is considered one of the indicators for improvement in clinical trials 20. For the diagnosis of NASH with advanced fibrosis or evaluation of steatosis in patients, liver biopsy is the gold standard. However, it is impossible to perform a liver biopsy in all patients with NAFLD that account for 25–30% of the general population 2121212121. Therefore, biomarkers for the noninvasive evaluation of NAFLD have been studied, such as serological and imaging biomarkers.8 Some biomarkers are based on a single parameter and the others are developed in the form of a panel that combines several parameters.7 Our study demonstrated that serum exosomal miRNAs showed significant differences according to the severity of NAFLD. NASH, an advanced form of NAFLD, is histologically characterized by inflammation, steatosis, and hepatocyte ballooning, and each histological feature is associated with different miRNAs. The diversity of miRNA expression profiles indicates that miRNAs are independently involved in the steatosis, inflammation, and ballooning process.
In addition, we identified correlations between laboratory data and expression of serum exosomal miRNAs. Alteration in the levels of liver enzymes, such as AST, ALT, ALP, and GGT, is important not only for the differential diagnosis of liver disease, but also for evaluating disease severity.22 However, no study has reported on the relationship between liver enzymes and miRNAs. In this study, we identified four miRNAs that are positively correlated with liver enzymes: miR-133b, miR-4436a, miR-4709-3p, and miR-8079.
Although the NAFLD Activity Score (NAS) is defined as the sum of steatosis, inflammation, and hepatocyte ballooning scores, when ANOVA analysis was performed, different miRNA expression patterns were found in NAS, steatosis, inflammation, and ballooning. However, in the Circos plot, the average miRNA expression level for each degree showed a similar pattern in each genomic region. Many miRNAs were observed corresponding to chromosomes 16, 17, and 19, which showed statistically significant differences based on the score, and fewer miRNAs were discovered corresponding to chromosomes 4 and 10. No miRNA correlated to the steatosis score was observed on chromosome 10. Interestingly, higher or lower correlation patterns between miRNA expression and two clinical parameters (ALT and PLT) were also detected with chromosomes 16, 17, and 19.
Functional network analysis between mRNA and miRNA is useful for evaluating the course of various diseases.23,24 We identified relationships between the miRNA expression patterns observed in this study and the gene expression patterns demonstrated in the GSE89632 dataset. In the GSE89632 dataset, subjects were divided into healthy controls and patients with steatosis and NASH, and demonstrated characteristics of fibrosis (stage), steatosis (%), lobular inflammation (severity), ballooning score, and NAS25. The study methodology was similar to our strategy. Hence, we used this dataset for the comprehensive network analysis.
When it comes to hepatic fibrosis, we found that low expression of hsa-let-7b-5p correlated with high expression of genes associated with liver fibrosis. The interaction between pro-fibrosis genes and hsa-let-7b-5p has been identified in interstitial pulmonary fibrosis and renal fibrosis.26,27 As for NAS score, low expression of miR-133b correlated with high expression of several genes. As miR-133b showed protective effects against allergic inflammation, 28 lower expression might increase hepatic inflammation. Taken together, the interaction between miRNA and mRNA is believed to play an important role in the progression of NASH.
The scarcity of publicly available gene expression data contributed to the limitation of our study. To the best of our knowledge, studies with similar approaches providing clinical parameters have not been previously performed, except for the GSE89632 study. The evidence demonstrated in our miRNA expression dataset needs to be validated in independent cohorts with larger sample sizes. The evidence and methods provided by us could be expanded to a comprehensive study of miRNAs and genes related to steatosis, fibrosis, ballooning, diabetes, and NAS. The small number of patients in each group constituting each variable is also a limitation. When ANOVA was performed with four variables, inflammation, steatosis, ballooning score, and NAS, it was difficult to ascertain a clear classification pattern by miRNAs describing each variable. Although the p-value indicating statistical significance was different for each variable, a minimum of 11 and a maximum of 25 miRNAs were identified that explained the classification of each variable. The data needs to be verified with larger patient cohorts in future studies.
We demonstrated differentially expressed miRNAs based on the groups of four variables with high, middle, and low values corresponding to the degree of severity (inflammation, steatosis, ballooning score, and NAS). Although NAS comprises three variables, different miRNAs have been identified, which are associated with each variable. Therefore, the pathways to explain each variable might be different. Comparative visualization was attempted with the construction of a Circos plot. MiRNAs that were significantly different with respect to each variable showed similar expression patterns in similar genomic regions. The integration of miRNA and mRNA expression analyses and network analysis also enabled us to interpret the differentially regulated genes from the perspective of systems biology in NASH. Using our methodology, co-expression networks of miRNAs and mRNAs could help reveal the pathological pathways of NAFLD, as well as provide new insight into several biological pathways related to liver functions. These pathways may also be used in the diagnosis of many liver diseases.