Impact of liver failure on the circulating extracellular vesicle miRNA repertoire

Cell‐derived small extracellular vesicles (sEVs) participate in cell–cell communication via the transfer of molecular cargo including selectively enriched microRNAs (miRNAs). Utilizing advances in sEV isolation and characterization, this study investigates the impact of liver injury and dysfunction on the circulating EV‐miRNA profile.


INTRODUCTION
through, for instance, the coordination of cellular activation, differentiation, and survival. 2,3 In addition to the studies exploring the intracellular functions of miRNA, particularly in the regulation of target genes, the relative stability of miRNA in biofluids has allowed detailed investigations into their role in mediating intercellular communication. The intercellular transfer of miRNA can be achieved through three fundamental mechanisms. The first is dependent on cell proximity and the formation of cytostructural bridging elements, including microtubules and gap junctions. The second involves extracellular release in association with stabilizing molecules, particularly the formation of protein-miRNA complexes with low-or high-density lipoproteins or with the Argonaute family of ribonucleoproteins. The third and best described method is encapsulation within extracellular vesicles (EVs), and this pathway is attributed with the majority of intercellular miRNA transfer. [4][5][6][7][8] EVs are a heterogeneous group of nanosized, membrane-bound particles, which are released by most cell types. The heterogeneity is in part attributed to varying modes of EV biogenesis. 9 Broadly, EVs that form upon outward budding of the cell membrane itself are known as microparticles, and these show greater morphological heterogeneity and can be up to 1000 nm in diameter. On the other hand, exosomes are approximately 100 nm in diameter, are formed within multivesicular bodies, and are characterized by proteins involved in this particular endocytic pathway, such as the tetraspanins CD9, CD63, and CD81. Exosomes have been extensively demonstrated to play a central role in intercellular communication by way of their varied cargo, which includes selectively enriched proteins, lipids, and nucleic acids, particularly miRNAs. They are shown to be stable in circulation and to be capable of targeting distant cells and organs. Owing to overlapping size, density and composition, coupled with imperfect isolation approaches, the description of EV isolates has recently shifted away from a focus on biogenesis and more toward a size-based description. Thus, isolates in the 50-200 nm range, which include exosomes, are currently referred to as small EVs (sEVs). 9 Circulating miRNAs have been proposed as diagnostic biomarkers in a variety of liver diseases, including nonalcoholic steatohepatitis, hepatocellular carcinoma, and cirrhosis, and in acute liver failure (ALF). [10][11][12][13] The majority of studies have performed analyses of total serum or plasma miRNA, which to some degree conflate the protein-and EV-associated fractions. The potential problems inherent in such undifferentiated analyses were highlighted in 2012 by Bala et al. in a study demonstrating that the underlying liver pathology influenced the miRNA repertoire of the two different compartments differently. 14 Thus, while the rise in circulating miR-122 and miR-155 observed in alcoholic and inflammatory liver disease was restricted to the sEV fraction, these same two miRNAs were predominantly increased in the protein-rich fractions during druginduced liver injury. Studies in different clinical contexts demonstrate similar divergence between the repertoire of miRNAs found in serum/plasma and that found in sEVs. These findings emphasize the need to study the sEV-miRNA compartment separately, given their selective enrichment and release from tissues and their potential for targeted transportation to distant sites.
In recent work, Blaya and colleagues performed comprehensive analysis of undifferentiated serum miRNA in patients with cirrhosis and ACLF. 15 Their data demonstrated miRNA to be profoundly dysregulated with worsening disease severity. miRNA profiles, however, exhibited significant overlap among healthy volunteers (HV) and those with compensated cirrhosis (CC), as well as between those with decompensated cirrhosis (DC) and ACLF. In their estimation, peripheral blood mononuclear cells were the most likely contributor to the dysregulation seen in the undifferentiated serum miRNA signature. As sEVs are selectively enriched in miRNA at the tissue of origin, we hypothesize that the sEV-miRNA fraction more closely represents changes within the liver and that this may enable clearer differentiation between groups of study.
Despite the significant advances in sEV isolation and characterization achieved since 2012, there has been a paucity of further investigations into the sEV-miRNA compartment in patients with liver injury and failure. Our current study combines advanced sEV isolation and characterization approaches with high-throughput miRNA analyses to investigate comprehensively the contribution of the degree of liver failure and the inflammatory response to the sEV-miRNA profile. To achieve this, a clinical cohort incorporating the spectrum of liver disorders, including ALF, compensated and decompensated chronic liver disease, and ACLF, was utilized and compared with both HV, but also to patients with severe sepsis but with preserved liver function. Particular attention was paid to ACLF given the paucity of biomarkers of prognostic value and the requirement for a better understanding of the molecular pathways involved in this condition. The results of this study demonstrate the dysregulation of circulating sEV-miRNAs with the increasing severity of liver disease and progression of cirrhosis. Furthermore, while cytokine profiles exhibit substantial overlap among individuals with ACLF and ALF, a distinct sEV-miRNA profile was noted in patients with ACLF with diagnostic and prognostic value.

Circulating cytokine profiling
Plasma cytokine analysis was performed using Meso Scale Discovery (MSD) V-Plex Human Proinflammatory Panel I. Plates and samples were prepared according to the manufacturer's protocols. In brief, 25 μL assay diluent was added into each well of a 96-well MSD plate.
The plate was then incubated for 30 min with vigorous shaking (300 rpm) at room temperature. Twenty-five μL of prepared standard (ranging from 0 to 2500 pg/mL) or plasma sample was dispensed into plate wells and incubated for 2 h with shaking at room temperature.
The plate was then washed three times with PBS (0.05% Tween-20).
Next, 25 μL of detection antibody solution was added into each well, the plate was sealed and incubated for a further 2 h with shaking at room temperature. After washing, 150 μL of 2 � read buffer T, a Trisbased buffer containing tripropylamine as a co-reactant for light generation, was added to each well. All samples were run in duplicate.
The plate was immediately read on the SECTOR(r) imager, and analysis was performed using DISCOVERY WORKBENCH 4.0 software according to the manufacturer's guidelines (MSD).

Small extracellular vesicle isolation and verification
sEVs, including exosomes, were isolated from platelet-poor plasma (PPP) by size-exclusion chromatography using CellGS Exo-Spin TM Mini Columns according to the manufacturer's instructions, and as previously described. 19 In brief, peripheral blood was drawn into BD Vacutainer ® K3-EDTA-coated collection tubes (Becton Dickinson), and processed to PPP by double centrifugation at 5000 and 10 000 g, respectively for 10 min. PPP was centrifuged at 16 000 g for 30 min. Data were processed with the nSolver 3.0 software (NanoString Technologies ® ).

Assay for miRNA functional target identification
To identify the target genes of selected miRNA we utilized the MISSION ® Target ID Library (Sigma) assay previously described. 20 This enables experimental identification of miRNA gene targets by screening the human transcriptome for functional binding sites. In so doing, this approach offers advantages over commonly used computational algorithms for miRNA target identification, which rely heavily on binding homology of miRNA strands and tend toward overestimation of putative targets, often identifying hundreds of potential targets, with the resultant lack of sensitivity or specificity evidenced from the paucity of experimentally validated miRNA targets. 20 In brief, the Target ID Library is a pool of plasmids, each with a human cDNA (16 923 unique genes) cloned downstream of a thymidine kinase-zeocin fusion protein (TKzeo) into Sfi1 sites in a p3 0 TKzeo vector, thereby conferring zeocin resistance and ganciclovir sensitivity (see Supporting Information S1: Supplementary Methods and Figure S1). This dual selection system allows cell transfection and the screening of the entire library at once, selecting first for stable transformants and secondly, after introducing a miRNA of interest, for functional cDNA targets. Selected cDNAs are then identified by sequencing. [21][22][23][24]

Statistical analyses
Unless otherwise stated, statistical analyses were performed using GraphPad Prism v7.0 Software. Categorical variables were compared with the chi-squared test. Continuous variables analyses were performed by Student's t-test, ANOVA, Mann-Whitney U or Kruskal-Wallis tests as indicated. Alpha error was adjusted according to Bonferroni correction was used to account for multiple comparisons.
Survival probability curves were calculated with the Kaplan-Meier method and compared with log-rank test. Results for continuous variables are expressed as median and interquartile range. Categorical variables are expressed as number and percentage (*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001). Further description of materials, methods, and miRNA statistical analyses can be found in Supporting Information S1. [25][26][27][28][29][30] Table 1. The predominant etiology of liver dysfunction in chronic conditions (CC, DC, ACLF) was alcohol excess, while in ALF was paracetamol/acetaminophen � mixed drug toxicity (Table S1). ACLF was associated with the poorest 90-day transplant-free survival among all groups studied at only 31%. The ACLF and ALF groups were associated with substantially and significantly higher liver disease severity scores, such as MELD and Child-Pugh, than all other groups studied, and exhibited significantly higher organ failure scores (SOFA). Inflammatory markers, including white cell count, and C-reactive protein were also higher among patients with ACLF than in CC, DC, and ALF, although these were highest among the SEP group as might be expected.

Multiplex analysis of serum circulating cytokine profile
Circulating plasma cytokine levels were analyzed across all patient groups ( Figure 1a). Patients with ACLF exhibited a significant difference from HV in seven of the 10 cytokines analyzed; a greater number than was seen among any other group of study. In keeping with previous descriptions, among these cytokines were inflammatory mediators such as IL6, TNFα, IL-8, IL-10, and IL-1, while IFNγ remained unchanged. 1,31,32 Further exploration of how circulating cytokine profiles compared among all groups of study was performed by principal components analysis (Figure 1b and Table S2). These data taken together highlight three key observations. First, there is significant overlap between patients with ACLF and those with ALF, despite the distinct onset, cause, and pathobiology of the liver dysfunction in these entities. Second, median cytokine levels progressively increased from healthy subjects to patients with CC, those with DC, and finally to those with ACLF or ALF, suggesting that disease progression and/or the degree of liver failure is associated with shifts in cytokine profiles. To determine whether the extent to

Degree of liver failure is the main determinant of small extracellular vesicle-miRNA profile
To gain further insight into the observed distinct sEV-miRNA profiles seen among patients with ACLF and ALF, we next performed cor-  Table 2 summarizes the number F I G U R E 1 Circulating cytokine levels exhibit overlapping profile among ACLF and ALF groups, which is distinct from other groups of study. (a) Individual log-transformed values of 10 plasma cytokines among all six groups of study. (b) Unsupervised principal components analysis of 10 circulating cytokines differentially expressed in at least one pairwise comparison exhibiting poorer demarcation between ALF and ACLF groups, but a distinct profile in patients with sepsis of nonabdominal etiology and without hepatic involvement (SEP). Comparisons between groups by Kruskal-Wallis test where *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. ACLF, acute-on-chronic liver failure; ALF, acute liver failure; CC, compensated cirrhosis; DC, decompensated cirrhosis; HV, healthy volunteers. To investigate further the salience of liver-derived sEVs on the circulating sEV profiles observed, we turned our attention to liverspecific miRNAs, of which miR-122 is the best known, making up as much as two-thirds of total miRNAs in the adult human liver. [33][34][35] In the cohort of recruited patients with ALF, miR-122 was the most significantly over-expressed miRNA in circulating sEVs as compared with HV (FC = 54.99, p < 0.0001, Figure 3b). While miR-122 was also more highly expressed among all other groups of patents with liver disease, and indeed among the CC group was again the most highly whereas green pixels indicate decreased miRNA levels. The profiles of ACLF and ALF sEV-miRNA exhibit distinct differences other groups. HEPATOLOGY RESEARCH overexpressed of the miRNAs (FC = 5.05, p < 0.001), this was to a significantly lesser degree than that seen in ALF (Figure 3c).

Four consensus miRNAs dysregulated in acute-onchronic liver failure
Given the poor prognosis associated with ACLF, we next turned our  Figure S3). sEV-miR-320e did not predict poorer survival outcomes in cohorts other than ACLF (data not included).
Based on these data, we turned to investigate the functional role of  Despite the overlap between ACLF and ALF in serum cytokine profiles, the two clinical conditions were different from the sEV-miRNA standpoint. Given our data suggesting that the sEV-miRNA fraction more closely represents changes within the liver, and given the stark difference in clinical characteristics between the two groups in terms of the presence or not of underlying liver disease, it is perhaps unsurprising that striking differences were noted in sEV-miRNA profiles in these groups. We identified four consensus sEV-miRNAs to be most discriminatory of ACLF: miR-320e, miR-374-5p, miR-202-3p, and miR-1910-5p. Of these, miR-320e showed significant association with poorer survival outcomes. These findings In summary, this study has demonstrated sEV-miRNA profiles to be significantly altered in progressive liver disease and has outlined miRNAs of diagnostic and prognostic potential in ACLF as well as establishing downstream gene targets. Further assessment of the F I G U R E 5 miR-320e regulates IK, RPS5, MANBAL, and PEBP1 expression. Cells containing Library constructs with miRNA target sites, and therefore surviving ganciclovir selection, were expanded. gDNA was isolated and PCR-amplified. (a) Before sequencing, Agilent TapeStation was used for the evaluation of the size distribution of purified gDNA, and (b) nanodrop performed for concentration and quality control. The resulting PCR product underwent sequencing, and a BLAST sequence search applied for miR-320e data displayed a high degree of homology for five transcripts: IK, RPS5, FEM1A, MANBAL, and PEBP1. The RNAhybrid computational algorithm determined the MFE of hybridization between miR-320e and the genes identified by sequencing. The lower the MFE, the higher the likelihood of miRNA targeting. (c) Example graphic of RNAhybrid computational output showing predicted hybridization between miR320e (green) and IK (Red) with predicted MFE indicated. (d) quantitative PCR analyses of the mRNA expression of identified targets were performed on Jurkat T cells transfected with either miR-320e mimic or inhibitor, where expression levels were relative to scrambled controls. gDNA, genomic DNA; MFE, minimum free energy.
HEPATOLOGY RESEARCH biological role of these sEV-miRNA should inform their contribution to the development of liver disease and ACLF.

ACKNOWLEDGMENTS
We would like to acknowledge the kind expertise, time, and support