Serum coding and non‐coding RNAs as biomarkers of NAFLD and fibrosis severity

Abstract Background & Aims In patients with non‐alcoholic fatty liver disease (NAFLD), liver biopsy is the gold standard to detect non‐alcoholic steatohepatitis (NASH) and stage liver fibrosis. We aimed to identify differentially expressed mRNAs and non‐coding RNAs in serum samples of biopsy‐diagnosed mild and severe NAFLD patients with respect to controls and to each other. Methods We first performed a whole transcriptome analysis through microarray (n = 12: four Control: CTRL; four mild NAFLD: NAS ≤ 4 F0; four severe NAFLD NAS ≥ 5 F3), followed by validation of selected transcripts through real‐time PCRs in an independent internal cohort of 88 subjects (63 NAFLD, 25 CTRL) and in an external cohort of 50 NAFLD patients. A similar analysis was also performed on liver biopsies and HepG2 cells exposed to oleate:palmitate or only palmitate (cellular model of NAFL/NASH) at intracellular/extracellular levels. Transcript correlation with histological/clinical data was also analysed. Results We identified several differentially expressed coding/non‐coding RNAs in each group of the study cohort. We validated the up‐regulation of UBE2V1, BNIP3L mRNAs, RP11‐128N14.5 lncRNA, TGFB2/TGFB2‐OT1 coding/lncRNA in patients with NAS ≥ 5 (vs NAS ≤ 4) and the up‐regulation of HBA2 mRNA, TGFB2/TGFB2‐OT1 coding/lncRNA in patients with Fibrosis stages = 3‐4 (vs F = 0‐2). In in vitro models: UBE2V1, RP11‐128N14.5 and TGFB2/TGFB2‐OT1 had an increasing expression trend ranging from CTRL to oleate:palmitate or only palmitate‐treated cells both at intracellular and extracellular level, while BNIP3L was up‐regulated only at extracellular level. UBE2V1, RP11‐128N14.5, TGFB2/TGFB2‐OT1 and HBA2 up‐regulation was also observed at histological level. UBE2V1, RP11‐128N14.5, BNIP3L and TGFB2/TGFB2‐OT1 correlated with histological/biochemical data. Combinations of TGFB2/TGFB2‐OT1 + Fibrosis Index based on the four factors (FIB‐4) showed an Area Under the Curve (AUC) of 0.891 (P = 3.00E‐06) or TGFB2/TGFB2‐OT1 + Fibroscan (AUC = 0.892, P = 2.00E‐06) improved the detection of F = 3‐4 with respect to F = 0‐2 fibrosis stages. Conclusions We identified specific serum coding/non‐coding RNA profiles in severe and mild NAFLD patients that possibly mirror the molecular mechanisms underlying NAFLD progression towards NASH/fibrosis. TGFB2/TGFB2‐OT1 detection improves FIB‐4/Fibroscan diagnostic performance for advanced fibrosis discrimination.


| BACKG ROU N D
Non-alcoholic fatty liver disease (NAFLD) accounts for the most increasing cause of chronic liver disease, hepatocellular carcinoma and of end-stage liver disease leading to liver transplantation. 1,2 Cohort studies demonstrated that among NAFLD populations, those with non-alcoholic steatohepatitis (NASH) have a higher risk of fibrosis progression, and that the presence of severe liver fibrosis is the main driver of hepatic and extra-hepatic prognosis. 2,3 Even if liver biopsy is considered the diagnostic gold standard, the availability of non-invasive markers to be used in NAFLD patients to predict NASH and/or severity of fibrosis represents a relevant medical need. 4 To date, the diagnosis of NASH in clinical practice by using non-invasive scores is difficult because of the lack of wellperforming and well-validated tools. 5 Furthermore, novel-proposed NASH blood biomarkers, such as CK18-Asp396 fragments, showed low sensitivity for NASH detection. 5,6 Different scores are available for the non-invasive assessment of fibrosis (eg 'aspartate aminotransferase (AST)-to-platelet ratio index' (APRI) and the 'four factors-based fibrosis index' (FIB-4); some of them are very easy to use and implement in clinical practice even if there are limitations such as false-positive results and high uncertainty areas. 3,7 When looking at imaging devices, liver stiffness measurement (LSM) through transient elastography (FibroScan) is a widely diffused tool, even if its accuracy can be influenced by obesity and severity of steatosis. 7,8 Circulating RNAs in plasma or serum have been attracting exponential attention as novel non-invasive diagnostic biomarkers of several kinds of diseases. Body fluid RNAs, including long non-coding RNAs (lncRNAs) and messenger RNAs (mRNAs), have many of the essential characteristics of good biomarkers, for example: (i) they are non-invasive accessible; (ii) their levels can easily be determined by basic molecular biology methods, first of all real-time PCR; (iii) they are remarkably stable in spite of the high amounts of endogenous ribonuclease in body fluids such as serum and plasma. [9][10][11] Several studies have highlighted the stability of lncRNAs and mRNAs, also under various experimental and pre-analytical oppressive conditions, including multiple freeze-thaw cycles, prolonged incubation at room temperature (up to 24h), exogenous RNAse treatments, time delay in processing of blood after venipuncture, low/high pH. 10,[12][13][14][15] The stability of lncRNAs and mRNAs in different body fluids against endogenous ribonuclease degradation may be provided by vesicles encapsulation and/or RNA binding protein association and also their highly stable secondary structures. 10,11,16,17 Two very recent studies reported the aberrant expression of long non-coding RNAs (lncRNAs) in liver tissue of NAFL/NASH patients and in mice models of NAFLD. 18,19 Furthermore, increasing trend ranging from CTRL to oleate:palmitate or only palmitate-treated cells both at intracellular and extracellular level, while BNIP3L was up-regulated only at extracellular level. UBE2V1, RP11-128N14.5, TGFB2/TGFB2-OT1 and HBA2 up-regulation was also observed at histological level. UBE2V1, RP11-128N14.5, BNIP3L and TGFB2/ TGFB2-OT1 correlated with histological/biochemical data. Combinations of TGFB2/

Conclusions:
We identified specific serum coding/non-coding RNA profiles in severe and mild NAFLD patients that possibly mirror the molecular mechanisms underlying NAFLD progression towards NASH/fibrosis. TGFB2/TGFB2-OT1 detection improves FIB-4/Fibroscan diagnostic performance for advanced fibrosis discrimination.

K E Y W O R D S
fibrosis, liquid-biopsy, NAFLD, NASH, RNAs

LAY SUMMARY
• Liver biopsy still remains the gold standard for NAFLD diagnosis confirmation, distinction between simple steatosis and NASH, and fibrosis staging. No study has been performed regarding the circulating lncRNA and mRNA signatures as biomarkers in NAFLD patient serum.
• Our study suggests the use of coding and non-coding RNA as non-invasive biomarkers of NAFLD and fibrosis severity. Combination of coding/non-coding RNA expression levels and clinical data could be conveniently applied as a diagnostic tool for non-invasive screening of NAFLD patients according to disease severity. experimental evidence is beginning to characterize the role of specific lncRNAs in the pathogenesis of liver fibrosis 20 and several metabolic functions, such as free fatty acid β-oxidation, lipogenesis and insulin secretion. 21 Numerous studies performed high-throughput analysis of mRNAs in liver tissues from NAFLD patients or in vivo animal models, which allowed the identification of several pathways associated with disease progression. [22][23][24][25][26] To date, no study has been performed on circulating mRNA and lncRNA signatures in sera from NAFLD patients. Therefore, the main aim of this work was to analyse the whole transcriptome profile in serum samples of NAFLD biopsydiagnosed patients to identify novel non-invasive biomarkers which are able to identify NAFLD patients with NASH and/or advanced fibrosis.

| Study subjects
This study involved two subject cohorts, namely: a first study cohort recruited in University Hospital of Palermo which included 71 patients with biopsy-proven NAFLD and 29 controls and a second external cohort recruited in University Hospital of Milan which included 50 NAFLD patients. A liver biopsy was performed to confirm the diagnosis of NAFLD in patients with ultrasound findings of fatty liver and/or persistent (>6 months) elevation of alanine aminotransferase (ALT) or AST. We excluded patients with alcohol-induced or drug-induced liver disease autoimmune or viral hepatitis and cholestatic or genetic liver disease from our study. We also excluded patients with current or past consumption of ethanol of more than 20 g per day. This study was conducted according to the Declaration of Helsinki and was approved by the ethics committee of the University Hospital of Palermo and Milan. Before the procedures, a written informed consent was obtained from all patients participating in this study. Further information regarding histopathological evaluation and sample processing can be found as Supplementary Data.

| RNA E X TR AC TI ON
We performed RNA extraction from serum samples of study cohorts, from liver biopsies of 12 subjects (four CTRL, four mild NAFLD NAS score ≤ 4 Fibrosis stage = 0, four severe NAFLD NAS ≥ 5 Fibrosis Stage = 3), and finally from HepG2 cells exposed to lipid treatment and their culture media. Total RNA was extracted by using miRNeasy mini kit (Qiagen). Further information regarding RNA extraction protocol can be found as Supplementary Data.

| MICROARR AY ANALYS IS
High-throughput profiling of serum coding/non-coding RNA, through microarray technology by Clariom D Pico Assay (Thermo Fisher Scientific), was performed on 12 study subjects: four mild NAFLD (NAS ≤ 4, Fibrosis stage = 0), four severe NAFLD (NAS ≥ 5, Fibrosis Stage = 3) and four CTRL. Further information regarding microarray experiments can be found as Supplementary Data.

| COMPUTATIONAL ANALYS IS
To understand the function of statistically significant deregulated transcripts for each comparison (mild NAFLD vs CTRL, severe NAFLD vs CTRL, severe vs mild NAFLD), we performed pathway enrichment analysis through Transcriptome Analysis Console v. 4 which retrieves canonical biological pathways from WikiPathways database and establishes P-values using two-sided Fisher's Exact Test (P < 0.05).

| IN VITRO CELL CULTURE E XPERIMENTS
Human hepatoblastoma cell line HepG2 was cultured as previously reported. 27 To obtain in vitro models of NAFL or NASH, HepG2 at 75% confluence were exposed to a mixture of oleate/palmitate or only palmitate at a final fatty acid concentration of 0.5 mM for 48 h respectively. 27 28

| S ING LE RE AL-TIME P CR A SSAYS
Validation of candidate coding/non-coding RNAs both in an independent internal cohort of 88 study subjects (63 NAFLD and 25 CTRL) and an in external cohort of 50 NAFLD patients was performed through real-time PCR assays (Power SYBR Green RNA-to-CT1-Step Kit Thermo Fisher Scientific) according to the manufacturer's protocol. Candidate transcripts were selected according to the following three criteria: (i) high levels of fluorescence microarray signals; (ii) low P-value in severe NAFLD vs mild NAFLD comparison (P ≤ 0.02); (iii) involvement in oxidative stress, inflammatory, apoptotic, authophagy or fibrogenic pathways according to the literature data. [29][30][31][32][33][34] Validated transcript expression was also analysed, through real-time PCR assays, in liver biopsies and in NAFL/ NASH in vitro models at intracellular and extracellular level.

| WHOLE TRANSCRIPTOME ANALYSIS
In order to identify novel potential biomarker signatures associated with NAFLD spectrum, we performed a whole transcriptome analysis in sera from four mild NAFLD patients (NAS ≤ 4; F = 0), four severe NAFLD patients (NAS ≥ 5; F = 3) and four healthy matched controls. Statistically significant deregulated transcripts in each comparison are represented as scatter plots in Figure 1A.

| PATHWAY ANALYS IS OF DIFFERENTIALLY E XPRE SS ED TR ANSCRIP TS
To explore the potential functions of deregulated transcripts, we identified significantly enriched biological pathways through Transcriptome Analysis Console. This bioinformatics analysis demonstrated that DE transcripts were statistically significant (Fisher's Exact Test P < 0.05) associated with several signalling pathways known to be involved in NASH/Fibrosis pathogenesis including inflammatory pathways (eg interleukins signalling pathways), metabolism deregulation (eg electron transport chain, TCA cycle nutrient utilization, Sterol Responsive Element Binding Protein, insulin signalling), cell death (eg apoptosis, ferroptosis), UPR stress (eg proteasome degradation, cytoplasmic ribosomal proteins), and extracellular matrix biosynthesis/fibrosis (eg collagen biosynthesis and modifying enzymes, VEGFA-VEGFR signalling pathway, EGF/EGFR signalling pathway) ( Figure 1B).

| VALIDATI ON BY qP CR s
In order to confirm microarray data, we analysed the expression of  Table 1. We selected the following transcripts: HBA2, UBE2V1, BNIP3L coding RNAs, RP11-128N14.5 lncRNA and TGFB2/TGFB2-OT1 coding/ non-coding RNA. We chose these transcripts because: (i) they presented high levels of fluorescence microarray signals; (ii) they presented a low P-value in severe NAFLD vs mild NAFLD comparison (P ≤ 0.02); (iii) they were potentially linked to oxidative stress, inflammatory, apoptotic, authophagy or fibrogenic pathways according to literature data. 29,30,35 We also selected one ncRNA GenBank

| VALIDATED TR AN SCRIP TS ARE A SSO CIATED WITH CLINI C AL MARK ER S AND HIS TOLOG IC AL SCORE S
We analysed whether -ΔCt values of validated transcripts were associated through a linear regression relationship both with the metabolic syndrome and liver damage-associated routine biochemical markers as with histological markers used in clinical practice to diagnose and stage NAFLD and fibrosis severity (Table   S2).
UBE2V1 expression levels correlated with AST, ALT, triglycerides, Kleiner ballooning and NAS scores ( Figure S8). BNIP3L expression was directly associated with Kleiner lobular inflammation, ballooning and fibrosis scores ( Figure S9). RP11-128N14.5 expression was associated with AST, Kleiner ballooning, NAS and fibrosis scores ( Figure S10). TGFB2/TGFB2-OT1 was associated through a positive linear regression relationship with FIB-4, Liver Stiffness Measurements, Kleiner lobular inflammation and fibrosis scores ( Figure S11). Finally, we found an inverse linear relationship among sherveebu AC020558.4 expression levels and triglycerides ( Figure S12) and between HBA2 and total cholesterol levels. P-values and R 2 values are reported in Table S2.

| E XPRE SS I ON ANALYS IS OF LIVER TISSUE S AND IN VITRO MODEL S
In order to further investigate the possible link between validated transcripts and cellular/molecular mechanisms associated with disease evolution, we evaluated whether RNA deregulation observed in serum of NAFLD patients was also present in liver tissue of NAFLD patients and in in vitro models of NAFL and NASH, both at intracellular and extracellular levels. Total RNA was extracted from liver bi-  Figure 4B). Furthermore, it is important to note that HBA2 was undetectable in HepG2 culture medium.

| RECEIVER OPER ATING CHAR AC TERIS TIC S CURVE ANALYS IS
By performing ROC curve analysis, we investigated, the di-   with AST or ALT in both cohorts decreased its AUC: 0.5 P = 1.

| D ISCUSS I ON
In this multiphase case-control study, we performed an initial screening of coding mRNAs and ncRNAs in biopsy-proven NAFLD patients at different stages of histopathological severity and controls (n = 12: 4 CTRL, 4 NAS ≤ 4 F = 0, 4 NAS ≥ 5 F = 3). Interestingly, performing computational analysis, we observed that deregulated transcripts, resulting from this initial screening, resulted to be statistically associated to key pathways known to be involved in fibro-  Several studies have highlighted the role of TGFB2-OT1 in the regulation of autophagy, apoptotic, inflammatory and fibrogenic pathways, all known to be involved in NAFLD progression. 43 It has been reported that TGFB2-OT1 is involved in the negative regulation of apoptosis and positive regulation of autophagy and inflammation in vascular endothelial cells. 31,44 Long Xiang Tu and colleagues reported the up-regulation of TGFB2-OT1 in fibroblasts derived from hypertrophic scars, suggesting its involvement in excessive proliferation of fibroblasts and accumulation of extracellular matrix. 45 As concerns TGFB2, its pro-fibrogenic role and/or its increased gene or protein expression has been reported in lens fibrosis, kidney fibrosis, cardiac fibrosis and liver fibrosis. 46 Several studies reported the involvement of BNIP3L in the regulation of pathways associated with NASH pathogenesis, specifically, apoptosis, necrosis, autophagy and fibrosis. Several studies supported BNIP3L involvement in apoptosis and necrosis of cardiomyocyte in heart diseases. 47 HIF1b-dependent BNIP3L increased gene and protein expression has been reported in primary cultured hepatocytes exposed to ethanol and in mice models of alcohol-induced liver injury and steatosis. 48 Although these data support the potential activation of this pathway in liver steatosis, we did not observe its activation in liver biopsies and in in vitro model at intracellular level. Concerning the pro-fibrogenic role of BNIP3L, Weili Liu and colleagues reported that the expression of BNIP3L is increased in cardiac fibrosis. 32 Although the major limitation of our study is the small sample size, our results may contribute to increase the understanding of RNAs should be assessed in liver biopsies of larger patient cohort.
Notwithstanding these limitations, our data contribute to expand the knowledge on molecular mechanisms associated with NAFLD pathophysiology. Moreover, they could provide the evidence for identification of novel biomarkers as future possible alternatives to liver biopsy.

ACK N OWLED G EM ENTS
We wish to thank the Scientific Bureau of the University of Catania for language support.

CO N FLI C T O F I NTE R E S T
The authors do not have any disclosures to report.