Bile extracellular vesicles from end-stage liver disease patients show altered microRNA content

Extracellular vesicles (EVs) have recently attracted attention as novel diagnostic biomarkers and therapeutic tools. Several reports have correlated blood EVs with liver diseases. However, blood EVs do not reflect the liver state as it contains other systemically circulating EVs. Therefore, we focused on bile EVs, which are secreted directly from the liver, for the identification of potential biomarkers of liver failure. Bile samples were collected from liver transplant recipients (n = 21) diagnosed with end-stage liver disease (ESLD) and donors (normal liver, NL; n = 18) during transplantation. Bile EVs were extracted using ultracentrifugation. Nanoparticle tracking analysis showed that bile EV concentration was significantly higher in recipients than in donors. Among recipients, bile EV concentration was remarkably higher in those with hepatocellular carcinoma. Next-generation sequencing revealed 461 and 465 types of microRNAs (miRNAs) in donor and recipient bile EVs, respectively, with no significant difference in diversity between the groups. Among 43 high-expression miRNAs, the expression of 86.0% of the miRNAs was higher in the bile EVs of recipients than in those of donors. Quantitative PCR validation showed that the levels of miR-17, miR-92a, miR-25, miR-423, and miR-451a significantly increased in bile EVs of recipients. Levels of miR-17 were remarkably higher in recipients with alcoholic ESLD. Secretion of EVs into the bile and their miRNA content increase in the ESLD state. Additionally, miRNA levels in bile EVs are not correlated with those in serum EVs. Bile EVs could be promising novel biomarkers for liver diseases.


Introduction
Liver cirrhosis (LC) is a terminal illness characterized by highly advanced liver fibrosis and caused by various chronic liver injuries. It can lead to hepatocellular carcinoma (HCC) and chronic liver failure, also called end-stage liver disease (ESLD) [1]. ESLD is reported as the 13th most common cause of death worldwide, increasing from about 67,600 in 1980 (1.54% of global deaths) to > 1 million (1.95%) in 2010 [2,3]. New effective treatments focusing on ameliorating liver fibrosis have been developed [4]; however, liver transplantation (LT) remains the only curative treatment for ESLD patients in clinical settings. The development of novel treatment options for LC requires further elucidation of the underlying molecular mechanism.
Several recent studies show that extracellular vesicles (EVs) are secreted by various cells into body fluids, such as blood, urine, and ascites, in normal and disease states [5]. EVs are small membrane vesicles and can be classified into exosomes, microvesicles, and apoptotic bodies, depending on the mechanism through which they are produced and their particle size. Exosomes are small vesicles (30-100 nm in diameter) enveloped by a lipid bilayer membrane containing functional molecules derived from secretory cells, such as microRNAs (miRNAs), messenger RNAs (mRNAs), and proteins that act as intercellular communication tools [5]. The profile of encapsulated molecules in exosomes varies depending on the conditions of secretory cells [6,7].
Several studies reported a relationship between EV characteristics and various disease states, including cases of liver disease [5,8]. In most liver disease studies, the characteristics of EV extracted from blood were determined, given that they are easy to collect. However, because EVs in circulating body fluids are composed of those secreted from different cells and organs, it is impossible to examine EVs originating from the liver alone in cases of liver disease. Therefore, herein, we focused on EVs in the bile, which are secreted from liver epithelial cells. Both hepatocytes and cholangiocytes release EVs into bile [9,10]. EVs extracted from bile are considered to reflect liver conditions more accurately than those found in the blood. Bile EVs are reportedly useful in the diagnosis of pancreatobiliary malignancies, including pancreatic adenocarcinoma and cholangiocarcinoma [11,12]. However, the relationship between bile EVs and liver diseases has not yet been unambiguously established.
In this study, we elucidated the characteristics of bile EVs from individuals with ESLD relative to those from individuals with normal liver (NL). Our results could lead to the identification of specific EVs associated with liver disease that might serve as novel diagnostic and treatment options for liver disease.

Materials and methods
Detailed experimental procedures are described in the Supplementary Material.

Patients and bile samples
From July 2015 to December 2016, 21 patients with ESLD who underwent living donor LT at Nagasaki University Hospital and 18 corresponding NL donors were enrolled in this study. This study was approved by the Ethics Committee of Nagasaki University (approval no. 15062234-2) and performed in accordance with the ethical principles of the Declaration of Helsinki. Written informed consent was obtained from all patients. Bile samples were collected from the gall bladder resected during LT as samples for ESLD; from the drainage tube inserted into the bile duct on the 7th and 14th day after LT; and from the resected gall bladder of NL donors. In three NL donors, bile samples could not be collected, because bile had been leaking from the resected gall bladder. To prevent decay of bile EVs and miRNA encapsulated in EVs, all bile samples were transported on ice and stored at − 80 °C shortly after collection during LT and thawed only once immediately before analysis. Serum samples were also collected from the patients just before LT and stored in a similar manner. The clinical characteristics and blood-examination data of recipients and donors were obtained from medical records.

Extraction of EVs from bile
EVs were extracted from 5 mL of bile in all cases. Bile samples were centrifuged at 300 g for 10 min at 4 °C, and the supernatant was centrifuged three times at 16,500 g for 10 min at 4 °C to remove cellular debris and floating cells. The supernatant was centrifuged twice at 150,000 g for 70 min at 4 °C to pellet EVs. The extracted EVs were resuspended in 200 μL phosphate-buffered saline (PBS) and stored in a low-adsorption tube at 4 °C until subsequent analysis. To confirm the quality and quantity of extracted EVs, morphological characteristics were examined using transmission electron microscopy (TEM; Kamakura Techno-Science, Inc., Tokyo, Japan). Concentration and particlesize distribution were analyzed using nanoparticle-tracking analysis (NTA; Theoria Science, Tokyo, Japan), and specific protein expression was analyzed using western blotting for CD63 and tumor-susceptibility gene 101 (TSG101).

Isolation of miRNA and next-generation sequencing (NGS)
After the addition of cel-miR-39 (0.01 ng/mL of bile sample) as spike-in for standardization, miRNA was extracted using the miRNeasy Mini kit (QIAGEN, Hilden, Germany) as per the manufacturer's instructions. Extracted miRNA was qualitatively and quantitatively evaluated using an Agilent 2100 bioanalyzer (Agilent Technologies, Foster City, CA, USA). A barcoded cDNA library was prepared using a SMARTer smRNA-Seq kit for Illumina (Clontech Laboratories, Mountain View, CA, USA) and sequenced using an Illumina HiSeq 2500 system (Illumina, San Diego, CA, USA) for comprehensive evaluation.

Statistical analysis
Results are presented as medians with a minimum-maximum range. For comparison between two groups, we performed a Mann-Whitney U test and non-parametric Spearman's correlation analysis using StatFlex (v.6.0; Artec, Osaka, Japan). Differences were considered statistically significant at p < 0.05. The Shannon index was analyzed by calculating the diversity index, H.

Patient characteristics
Bile samples were collected from 21 LT ESLD patients and 18 NL donors. Patient characteristics are shown in Table 1.

3
All LT recipients had ESLD, and the background liver diseases were hepatitis C virus (HCV; n = 7), alcohol-related (= 5), non-alcoholic steatohepatitis (NASH; n = 3), autoimmune hepatitis (AIH; n = 2), and others (n = 4). HCC was complicated in five patients (23.8%), and tumor-nodemetastasis (TNM) stages and incidences were as follows: stage I, two cases (40%); and stage II, three cases (60%). Comparison of baseline characteristics between recipients complicated with and without HCC is shown in Supplementary Table 1. The median Child-Pugh score and model for ESLD score were 11 and 19, respectively. Mild fatty liver was observed in four NL donors (22.2%). Bile samples obtained from donors were used as healthy controls for comparison with ESLD.

Characteristics of bile-extracted EVs
After extraction, the morphological characteristics of EVs were observed using TEM. Bile EVs were round and ranged from 30-100 nm in diameter, which was similar to that reported previously (Fig. 1a) [13]. Subsequently, analysis of the concentration, size, and distribution of bile EVs by NTA revealed a concentration between 1.60 × 10 10 and 26.4 × 10 10 particles/mL, with a particle-size peak at ~ 100 nm (Fig. 1b). Furthermore, western blotting identified levels of CD63 and TSG101, as key EV protein markers commonly expressed in exosome populations, in the extracted bile EVs (Fig. 1c). These results suggested that EVs extracted from bile were rich in exosomes.

Bile EV concentration increased in liver disease
NTA revealed that the concentration of bile EVs was significantly higher in ESLD patients than in NL donors (p = 0.008) (median concentration of bile EVs: ESLD patients, 14.14 × 10 10 ESLD vs. 5.45 × 10 10 particles/mL, NL donors) (Fig. 2a). After LT, the concentration of bile EVs decreased to 4.66 × 10 10 particles/mL (n = 17) and 2.78 × 10 10 particles/ mL (n = 15) after 7 and 14 days, respectively, with these values significantly lower than those obtained before LT (before LT vs. 7 days after LT: p < 0.001; before LT vs. 14 days after LT: p < 0.001) (Fig. 2b). There was no significant difference in the concentrations of bile EVs between patient groups classified according to background disease-causing ESLD (Fig. 2c). However, the concentration of bile EVs in patients with HCC was 22.89 × 10 10 particles/mL, which was significantly higher than that in patients without HCC (10.45 × 10 10 particles/mL) (p = 0.026) (Fig. 2d). The differences in EV concentrations based on the TNM stages of HCC were not statistically evaluated owing to the small number of cases (data not shown). The particle size of bile EVs was also examined similarly, revealing no significant differences in the particle size of bile EVs between ESLD patients and NL donors or between groups of patients classified according to clinical characteristics ( Supplementary Fig. S1).  Fig. S2). To identify miRNAs specifically contained in bile EVs of ESLD patients, we extracted 43 types of miRNAs detected at a median of ≥ 5 reads per million (RPM) in bile EVs of either the ESLD patients or NL donors (Fig. 3). Of these, 38 types (88.4%) of miRNAs were found at higher levels in bile EVs of ESLD patients, 12 of which were significantly higher than those in bile EVs of NL donors (Supplementary Table 3). No miRNA was found at a significantly higher level in the NL donors than in ESLD patients.

Quantification of miRNA in bile EVs extracted from ESLD patients using real-time quantitative PCR (qPCR)
NGS analysis revealed a total of six miRNAs, including miR-17, miR-92a, miR-25, miR-423, and miR-451a, at significantly high levels in bile EVs of ESLD patients; and liver-specific miR-122; which were validated using qPCR. Similar to NGS results, there was no significant difference in miR-122 levels in bile EVs between the two groups (p = 0.949); however, levels of miR-17, miR-92a, miR-25, miR-423, and miR-451a were significantly higher in ESLD patients than in NL donors (p = 0.002, p < 0.001, p < 0.001, p = 0.005, and p = 0.001, respectively) (Fig. 4a). Comparison of patients grouped together based on background liver disease indicated that miR-17 expression was significantly higher in alcohol-induced ESLD than in other background diseases (HCV vs. alcoholic: p = 0.012; alcoholic vs. NASH: p = 0.042) (Fig. 4b). Conversely, HCC did not affect miRNA levels in bile EVs. ( Supplementary  Fig. 3). To determine whether changes in miRNA-expression levels in bile EVs were similar to those in serum, levels of these six miRNAs were measured using qPCR using serum samples from each patient. We observed no significant correlation between changes in expression levels of EVs in bile and serum in these miRNAs, indicating that the miRNA profile in bile EVs was not reflected in serum (Fig. 5).

Discussion
We compared bile EVs in ESLD patients who received LT with those of donors with NL to identify potential ESLD biomarkers. We found that levels of EVs secreted into the bile increased in ESLD patients, particularly in patients with HCC. Second, there were no major changes in the overall diversity of miRNAs in bile EVs between the two study groups, although the total amount of miRNA was higher, and the types of miRNAs changed considerably in the bile EVs of ESLD patients. Third, we found that altered miRNA expression in bile EVs was not reflected in serum miRNA EVs from patients with ESLD. Levels of miRNAs in bile EVs were analyzed by real-time qPCR, normalized against that of cel-miR-39, and corrected by each EV-particle concentration to examine miRNA level per EV. Outliers were removed from the scatter plot. a Levels of miR-17, miR-92a, miR-25, miR-423, and miR-451a significantly increased in patients with ESLD (recipients, n = 21) relative to those in NL (donors, n = 16). miR-122 level did not differ significantly between the two groups. b The level of miR-17 was significantly higher in alcoholic ESLD (n = 5) than in ESLD due to HCV (n = 7) and NASH (n = 3). miR-92a level was significantly higher in alcoholic ESLD than in ESLD due to HCV. There was no significant difference in other miRNAs levels between HCV-, alcohol-, and NASHrelated ESLD. Mann-Whitney U test and non-parametric Spearman's correlation analysis were used to determine statistical significance. *p < 0.05; **p < 0.01; ***p < 0.001. N.S. not significant from ESLD patients. Furthermore, to the best of our knowledge, this study is the first to analyze bile EVs extracted from NL (healthy control). EVs are secreted by several cells and vary in count and content depending on the type or health of the EV-producing cells. Thus, they have become an area of intense research as potential diagnostic biomarkers for various diseases [6,7]. A majority of these studies have focused on EVs extracted from blood samples. The relative ease of the extraction process has made EVs a promising, clinically effective biomarker. However, EVs extracted from blood include EVs secreted from various organs; therefore, blood EVs only indirectly reflect changes in EVs secreted from a certain type of cell or tissue. Importantly, it is not completely reflective of changes in the EV secretion profiles of hepatocytes in the state of liver disease [14]. By contrast, bile EVs are secreted into the bile canaliculi, which do not interact with the sinusoids. Thus, the only cells that secrete EVs into bile canaliculi are hepatocytes and bile duct cells [10]. Therefore, bile does not contain EVs secreted from other organs or from nonparenchymal liver cells, such as hepatic stellate cells, Kupffer cells, and sinusoidal endothelial cells, indicating that changes in bile EVs more accurately reflect changes in EVs secreted from hepatocytes, which are exposed to the liver microenvironment as compared with blood EVs. Indeed, many studies have reported the altered expression of various miRNAs in blood and liver tissue of LC patients [15][16][17]. However, the miRNAs in bile EVs identified in this study differed greatly from their results, suggesting that bile EVs contain a unique miRNA profile independent of the blood and liver tissue samples.
Bile EV concentration was significantly higher in ESLD liver than in NL, as revealed using NTA. Furthermore, bile EV concentration decreased promptly after transplantation with NL, reaching levels similar to those in NL donors, suggesting that EV secretion by hepatocytes increased following exposure to the liver microenvironment during ESLD. We found no associations between bile EV concentration and underlying disease, alanine aminotransferase (ALT) levels, or hepatic reserve (data not shown). Previous findings indicate that blood EV concentration increases with liver injury [18,19]. Moreover, changes in blood EV concentrations correlate with ALT, hepatic fibrosis, cell death, and pathological angiogenesis. In mouse models of nonalcoholic fatty liver disease, blood EV concentration increased early on after liver injury and subsequently displayed time-dependent Fig. 5 Comparison of miR-17, -92a, -25, -423, -451a, and -122 levels between EVs from bile and whole serum. Levels of miRNA in EVs in bile and whole serum were analyzed by real-time qPCR. miRNA levels in bile EVs were normalized against that of cel-miR-39 and corrected by the concentration of EV particles in order to determine miRNA content per EV. Serum miRNA levels were normalized against that of cel-miR-39. No significant correlation was observed between any of the miRNAs and the source of EVs. Mann-Whitney U test and nonparametric Spearman's correlation analysis were used to determine statistical significance changes [18]. These studies show that EV secretion increases several folds, although these likely reflect not only the EVs secreted by hepatocytes but also systemic changes, including EV secretion by other organs that accompany fibrosis progression.
Patients with ESLD complicated with HCC had significantly higher EV concentrations than those without HCC. A recent study reported the superiority of bile EVs for diagnosing malignant bile duct stenosis as compared with tumor markers, such as serum Ca19-9 [11]; thus, bile EV concentration might also be effective in diagnosing HCC. EV secretion is more elevated in cancer cells than in normal cells, and EVs from cancer cells are involved in angiogenesis, inflammation, cancer-related fibroblast differentiation, and epithelial-mesenchymal transition in the tumor microenvironment [20,21]. Increases in EV secretion by HCC itself might result in higher bile EV concentrations; however, the pathway of EV secretion into the bile duct by HCC cells that have lost polarity and normal function remains unknown, thus warranting further investigations.
Here, we found that the expression of five bile EV miR-NAs (miR-17, miR-92a, miR-25, miR-423, and miR-451a) significantly increased in the ESLD liver relative to NL. Furthermore, a comparison of underlying disease or presence of HCC showed that miR-17 expression significantly increased in alcohol-induced ESLD, which is consistent with a report of markedly higher miR-17 and miR-18a levels in the liver tissue of a rodent model of alcohol-induced liver injury [22]. miR-17 and miR-92a are components of the miR-17-92 cluster, and miR-25 comprises the miR-106b-25 cluster, both of which are inhibited by the transcription factor E2F and represent oncogenic miRNAs [23,24]. It is possible that these EVs secreted by hepatocytes in the microenvironment of liver injury with fibrosis progression contribute to HCC progression and could be promising targets of new liver cancer treatments. In terms of the function of EVs released into the bile, analysis of a rat model indicated that EV uptake occurs via cholangiocyte cilia to modulate bile duct cell growth [10]; however, determination of those target cells or how they functionally transmit their content remains unknown. The present study did not elucidate how EVs released by hepatocytes in the liver injury microenvironment act on bile duct cells or whether they also act on the gastrointestinal tract after secretion. Future studies should investigate these aspects in more detail.
These results must be interpreted with caution, and some limitations should be noted. First, because the study participants were recipients undergoing living-donor LT, the sample size was small. Therefore, statistical analysis on the background factors of liver failure (presence of background liver disease and hepatocellular carcinoma) was insufficient. More cases are needed to clarify the effects of HCC progression and other background liver diseases on bile EVs. Second, we considered that changes in EV content were caused by environmental changes in hepatocytes rather than cholangiocytes. It is difficult to separate these using in vivo samples because specific EV markers for distinguishing them have not been identified. However, it is possible that cholangiocytes might be involved in changes in EV content. Third, in the analysis of miRNA-expression levels, exogenous cel-miR-39 spike-in was used for normalization. However, the deviation of miR-39 levels in bile EVs was large between individual cases (a few samples showed outliers in the scatter plot). Li et al. [12] similarly reported variations in the level of the same amount of miR-39 added as an internal control in bile upon qPCR quantification, suggesting that the efficiency of RNA extraction from bile was not constant among the samples [12]. Therefore, further verification of the relationship between miRNA levels and disease state is needed. Additionally, the establishment of a reliable internal control for bile samples will help improve future studies. Furthermore, collecting bile samples from chronic liver disease patients is invasive and difficult; thus, it may have limited utility as a clinical biomarker. By contrast, bile samples can be obtained easily via drainage tubes placed in patients after LT. Future studies should evaluate the benefits of bile EVs as a diagnostic tool of post-transplant complications, such as acute rejection, blood-flow obstruction, and bile duct stricture. Lastly, the practical utility of our study is the finding of increased numbers of EVs in ESLD and HCC patients. Further examination of bile EVs is required for verification of potential clinical applications.
In conclusion, EVs secreted from the liver into the bile change to reflect liver disease state. Furthermore, miRNAs in bile EVs change as they do in blood; however, the changes are independent of changes in serum miRNA. Various studies have already focused on EVs in blood samples, but in this study, we found that EVs in bile can also serve as new biomarkers that provide direct insights into the liver microenvironment in liver disease.