A total of 490 consecutive subjects belonging to a single ethnic group (Bengali) with history of significant ethanol consumption (more than 80 g/day for more than 10 years) (16, 19) was enrolled for the study from the indoor and outdoor of Department of Hepatology at IPGME&R, Kolkata, East India and Department of Medicine at AGMC, Tripura, North-East India. Diagnosis of ALC (n=168) was based on ultra-sonographic evidence of normal liver with normal level of ALT/AST, while cirrhotic ALD patients (n=322) were diagnosed with portal hypertension, esophageal/gastric varices with/without ascites and steatohepatitis patients with fatty liver, elevated level of ALT but no histological sign of cirrhosis. Following the same criteria, 100 ALD patients were included in the validation cohort. Healthy individuals (n=10) having normal ALT/AST were also included.

Although samples were included from two different regions of India having similar ethnicity, both the groups exhibited same pattern of allelic distribution, hence samples were clubbed.

Subjects from different ethnic background, positive for anti-HBsAg, Anti-HCV, or HIV seropositive, having co-morbidities like diabetes mellitus, and unwilling to comply with the study protocol were excluded.

Blood (5ml) was collected from every participant in EDTA-vial and used for isolation of chromosomal DNA (32), PBMC and neutrophil using histopaque 1077 (Sigma). Demographic, clinical and biochemical data of non-alcoholic normal, ALC and ALD patients are presented in Tables 1A and 1B .

Non-alcoholic steatohepatitis (NASH) patients (n=50) with comparable age, sex and having no other liver diseases as verified histologically were included from the Department of Hepatology, IPGME&R, Kolkata, India ( Supplementary Table 1 ).

Explanted liver tissues were collected from patients undergone orthotopic liver transplantation due to alcoholic cirrhosis (n=4) from Centre for Liver and Biliary Sciences, Indraprastha Apollo Hospital, New Delhi, India and confirmed with histological, and radiological evidences.

Normal liver tissues (n=10) were collected from gastrointestinal surgery clinic of IPGME&R, Kolkata, during cholecystectomy of gall bladder carcinoma (GBC) patients as routine analysis to test liver metastasis. Liver tissues with normal architecture (n=4) were included in the study after histological verification. Tissues were collected in RNA later (ThermoFisher, Cat. #AM7020) and preserved at -80⁰C for future use.

RNAseq data of alcoholic hepatitis (n=5) vs. normal liver tissue (n=5) (GSE143318) was retrieved from https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE143318 and the data was analyzed using DEseq2 package to obtain differential gene expression (DEG). DEG with log₂fold change>2 and P_adjusted value <0.05 were considered for further pathway analysis using Kyoto Encyclopedia of Genes and Genomes (KEGG).

A BioPlex ProHuman chemokine Screening Panel, 48 plex (Bio-Rad, # 120072831) was used to detect the levels of forty serum chemokines following the manufacturer’s protocol at National Institute of Biomedical Genomics, Kalyani, India. The data were analyzed using BioPlex® 200 Multiplexing Platform. DGE were performed as mentioned earlier and subjected to KEGG pathway analysis.

The total promoter sequence of TNF-α and IL1ß were obtained from the University of California Santa Cruz (UCSC) Genome Browser, selecting the Human GRCh37/hg19 genome assembly. The entire promoter sequence was scanned using TFBIND tool for the transcription factor binding site. ChIP Seq (https://genome.ucsc.edu/) data was used for validation.

Liver tissue (1mg) was homogenized and RNA was isolated using Trizol (ThermoFisher, # 15596026). About 2.5μg/500 ng of total RNA was used to synthesize cDNA for mRNA/miRNA using Revertaid Reverse Transcriptase (ThermoFisher, #MAN0012885)/miSCRIPT PCR starter kit (Qiagen, #218193) respectively. Expression of immune related genes was assessed in ABI QuantStudio 7 real-time PCR machine using SYBR green (Roche). Only upregulated genes in each sample were further considered for genotyping. Primers used for quantification of mRNAs and miRNAs are presented in Supplementary Table 2A .

Chromosomal DNA was isolated using salting out method from 3ml of EDTA-blood (14, 32). Variations in the promoter région was determined by sequencing with Big-Dye Terminator v3.1 (ThermoFisher, #4337455) on ABI Prism 3100 Genetic Analyzer followed by multi-sequence alignment using Sequencher software.

HepG2 (Hepatoblastoma), and LX2 (Stellate cell) cell lines were maintained in DMEM (HiMedia # AL067) with 10% and 2% FBS respectively while THP1 (monocytic cell line) was maintained in RPMI with 10% FBS. Endothelial cell growth media (Cell applications Inc.#211-500) was used for HUVEC (primary Endothelial cell line, Invitrogen). Cells were maintained at 37°C incubator with 5% CO₂.

The full-length (FL) TNF-α cDNA was cloned in pcDNA3.1(+)/myc-HisB vector (ThermoFisher). Dr. Susanta Roychowdhury, Indian Institute of Chemical Biology, Kolkata, India gifted us two plasmids carrying FL-cDNA of the IL1β gene with promoter having -31CC/TT (pMC-31CIL-1β/pMC-31TIL-1β) (33). Premir-124 was cloned in pRNAT-U6.1/neo vector and pSiCHECK-2 dual-luciferase reporter vector was used for the cloning of 3’UTR region of each gene. The site-directed mutagenesis (SDM) kit (NEB) was used to generate miRNA binding site mutant following manufacturer’s protocol. Primers for 3’UTR cloning and SDM are presented in Supplementary Tables 2B, C .

Lipofectamine 2000 (ThermoFisher, #11668019) and x-tremeGENE HP transfection reagents (Sigma, #6366236001) were used for transfection in different cell lines as required. THP1 cells were treated with LPS (100ng/ml) for 12 hours after transfection. HepG2/THP1/LX2/HUVEC were either directly transfected with required plasmids or incubated with conditioned media (CM) of transfected THP1 cells (1:1 ratio) for 48 hrs. Each experiment was performed in duplicate and repeated three times.

Secreted levels of TNF-α/IL1β in the serum of ALD/ALC individuals, and in cell culture supernatant were determined using ELISA following the manufacturer’s protocol (Ray BioTech, #P01375 and Abclonal, #RK00001).

About 2x10⁶ PBMCs were seeded on 6 well plates. After 24 hours, cells were either stimulated with LPS (100ng/ml) for 6 hours or left untreated and subsequently total RNA was isolated.

Apoptotic cells were quantified in BD-FACS VERSE using Annexin V and PI staining kit (Invitrogen, #556547) after 48 hours of transfection following manufacturers’ protocol. Early (Annexin v-FITC⁺) and late apoptotic cells (Annexin V-FITC⁺PI⁺) were considered for quantification of total apoptotic cells.

HUVEC cells (5x10⁴ cells/well) were seeded on the collagen pre-coated upper well of the Boyden Chamber (Cell BioLabs, 3μm pore size, #CBA-103) and either directly transfected or conditioned media (CM) of pTNF-α/Anti-TNF-α anti-miR-124-3p oligo transfected in THP1 cells (1:1) were added. The lower chamber was filled with 30% FBS containing media and freshly isolated neutrophil (10⁶ cells) was added on the upper chamber and number of cells migrated to the lower chamber was counted at different time points (2-24 hrs) using hemocytometer.

All the statistical tests were performed using SPSS version 10.0 and Graphpad prism. Hardy-Weinberg Equilibrium test was done for each SNP to determine whether the SNP present in this population is in equilibrium or influenced by any evolutionary forces. The association of genotype and allele frequencies were calculated using Pearson’s chi-square test (14, 19). Unpaired two tailed Student’s t-test was performed for comparison of expression data. T-test and Mann Whitney U test was employed for biochemical parameters as required. The parameters were presented either as mean ± SD (for normally distributed parameters) or as median and minimum to maximum distribution with range of the datasets (for skewed distributed parameters). Individual predictors were determined using multiple logistic regression analysis with significantly associated SNPs in Univariate analysis. P-values were corrected using Benjamini−Hochberg multiple testing correction. P_adjusted <0.05 was considered as significant.

This study has included 490 individuals in discovery cohort (ALD=322 and ALC=168) and 150 individual in validation cohort (ALD=100 and NASH=50). In addition, 10 healthy individuals were incorporated. Explanted Liver tissues were also collected from four ALD patients. Liver biopsies from gall bladder carcinoma patient (n=10) with normal histology were considered as normal (n=4).

Despite having similar amount of alcohol consumption, all alcoholics are not susceptible to ALD suggesting genetic variations play crucial role here. To identify the genes responsible for this differential behavior, overexpressed genes in ALD were identified after analysis of the transcriptome data, GSE 143318 (Alcoholic hepatitis vs. Control). It was followed by KEGG pathway analysis and revealed that the chemokine signaling pathways were the most upregulated pathways in ALD ( Figure 1A ). Furthermore, our Bio-Plex Pro™ Human Chemokine Panel, 40-Plex assay with serum of ALD (n=28) and ALC (n=20) individuals also disclosed cytokine-cytokine receptor signaling and chemokine signaling pathways were the highest upregulated pathways in ALD ( Figure 1B ). Thus, we focused on upregulated immune genes only and genes were selected after validation in the liver tissue of ALD (n=4) and ALC (n=4) using customized qRT-PCR expression array which included immune genes associated to alcohol induced steatohepatitis and cirrhosis such as TLR4, IFN-α, CD14, MCP1, IL8, TGF-β, TNF-α, IL6, IL18, IFN-γ, IL22, IL17, IL10, CCR2 and complements C5 and C3 as found in the published literatures (19, 34, 35) and in the above two analyses. After comparative analysis of the above two arrays, six genes TNF-α, IL6, MCP1, IL1β (P value<0.05) and TGF-β, IL8 (border line significant in liver tissue) were found up-regulated in both the serum and liver tissue samples of all the ALD patients tested compared to ALC ( Figures 1C, D ), which suggested their significant contribution in the ALD development.

Now, to find the reason of ALD development in only 10-30% of chronic alcoholics, the serum cytokine data of all the six genes in ALC group was carefully investigated ( Figure 1C ) and observed that this group had significantly lower level of serum cytokines albeit they were exposed to the same volume of alcohol compared to the ALD patients. A group of ALC individuals were noted with cytokines level below the detection limit. To understand this disparity in expression, the entire promoter regions of all the above six genes were sequenced in 322 ALD and 168 ALC individuals and noticed no sequence variations in MCP1, IL6, TGF-β and IL8 promoters. But, three SNPs, rs361525 (G/A) at -238 of TNF-α promoter and rs1143627(C/T) at -31 & rs16944(C/T) at -511 in IL1β promoter were noted and the frequencies were more in ALD individuals than ALCs (15.9% vs.3%, p<0.01; 65.5% vs. 40%, p<0.01; 48% vs. 61.8%, p<0.01 respectively) ( Table 2A ). Further univariate analysis showed significant association of all the three SNPs with ALD but multivariate analysis revealed that rs361525G/A [(OR 1.75, 95% CI (1.22-2.76)] and rs1143627/C/T+TT (OR 2.32, 95% CI (1.78-3.62) were two independent risk alleles which might confer susceptibility to ALD ( Table 2B ). This data was again validated in a new cohort of 100 ALD patients and 50 NASH patients and insignificant association was observed with NASH patients (Data not shown).

Now, to determine the impact of two SNPs on the expression of TNF-α and IL1β, normal individuals carrying the above SNPs were tested by qRT-PCR and ELISA. To our surprise, the expression of TNF-α and IL1β were noted significantly higher in the liver tissue of the normal individuals carrying rs361525GA at -238 (n=6) than GG (n=9) and rs1143627CT at -31(n=9) than CC (n=6) genotype respectively ( Figures 2A, B ). ELISA data of normal individuals with risk allele rs361525GA and rs1143627CT also showed higher level of TNF-α and IL1β than GG and CC allele respectively ( Figures 2C, D ). These data suggest that aforementioned SNPs in the promoter of TNF-α and IL1β might be contributed to the disease pathogenesis in susceptible individuals.

Next, to understand the mechanism of overexpression of TNF-α and IL1β in susceptible individual, UCSC-ChIP-seq data was analyzed and observed a NFκβ binding site at -239 position in the TNF-α promoter. Thus, Luciferase reporter assay with the TNF-α promoter constructs carrying either -238GG or -238AA genotype was performed in presence or absence of LPS and p65-NFκβ subunit. It was observed that the genotype “AA” showed higher luciferase activity than GG ( Figures 2E, F ) indicating that the presence of “A” at -238 loci might enhance the binding affinity of NFκβ to that site. Similarly, PBMC isolated from normal individual having either -31C/C or -31C/T allele in IL1β and subsequent qRT-PCR data depicted that the individual with -31C/T allele had significantly higher expression of IL1β than -31C/C at IL1β loci (T/T allele was absent in this population), in both presence or absence of LPS ( Figure 2G ). Next, when HepG2 cells were transfected with either pIL1β-31TT or pIL1β-31CC plasmids, the expression of IL1β was noted higher in presence pIL1β-31TT than pIL1β-31CC by qRTPCR and ELISA as well ( Figure 2H ). It is worthy to be noted that a new TATA box appears when the C allele replaces with T allele at -31 loci of IL1β promoter. So, binding of more TATA box binding protein plausibly induced the expression of IL1β in CT genotype carrying individuals.

In 2000, Hill et al, have reported that alcohol/LPS stimulates the expression of the nuclear factor NFκβ, which induces synthesis of TNF-α (36) and TNF-α and IL1β trigger a wide range of intracellular signaling pathways, which require IL6, MCP1, ICAM1 and TGF-β. To our surprise, multiple NFκβ binding sites were noted in the promoter regions of all these six genes which were significantly altered in our cohort ( Supplementary Figure 1A ). The data was validated in THP1 cells treated with LPS and observed overexpression of TNF-α, IL1β, MCP1, IL8 and NFκβ while the expression of each gene was also synergistically increased upon overexpression of TNF-α in HepG2 cells treated with LPS ( Supplementary Figures 1B, C ).

Thus, our data suggest that LPS induces expression of TNF-α and/or IL1β and individual carrying either rs361525G/A in TNF-α or rs1143627C/T in IL1β produces excessive amount of TNF-α and IL1β in the liver milieu. Now, TNF-α signaling through TLR4 pathway activates multiple intracellular as well as intercellular signaling cascades that lead to progression of liver diseases from steatosis to steatohepatitis, fibrosis and cirrhosis (13).

In purpose to abolish the effect of increased level of TNF-α and IL1β upon alcohol exposure, extensive bioinformatics analysis was performed using Targetscan, miRDB and miRwalk to select a common miRNA targeting both the genes. Here, a few more selection criteria were employed such as the selected miRNA was not abundant in the liver, and it was unaltered upon alcohol exposure. Following above criteria, miR-124-3p was selected which was an anti-inflammatory miRNA. The read count of miR-124-3p was evaluated from next generation sequencing data of ALD vs. normal liver tissue and it was quite low (Read Count<5, data not shown).

Next, the data obtained in bioinformatics analysis was validated by 3’-UTR-luciferase reporter assays using both wildtype (WT) and miR binding site mutant (MT) 3’UTR-Luciferase constructs. Thus, HepG2 cells were co-transfected with premiR-124 and WT/MT 3’UTR-luciferase constructs of TNF-α and IL1β. After 48h of transfection, a significant reduction in the luciferase activity was observed with WT construct and with increasing doses of miR-124-3p while the MT-luciferase constructs showed no response to miR-124-3p. Furthermore, ELISA data with the conditioned media (CM) of miR-124-3p overexpressed HepG2 cells depicted that the restoration of miR-124-3p in HepG2 suppressed the level of TNF-α and IL1β by 2.5 folds and 4 folds respectively ( Figures 3A–F ). So, these data clearly suggested that the binding of miR-124-3p to the 3’UTR of TNF-α and IL1β.

Upon alcohol consumption, the gut derived bacterial endotoxin LPS mediates inflammation through CD14 and TLR4 to initiate MyD88 dependent and TRIF signaling cascade to activate NF-κB which regulates TNF-α and IL1β secretion from activated Kupffer Cells (KCs) (11). TNF-α binds to the TNF-α receptor (TNFR1) on the hepatocytes and activates TNFα receptor-associated death domain (TRADD) mediated extrinsic apoptosis pathway (37). Here, HepG2 cells were incubated with the CM of the LPS treated THP1 cells transfected with pTNF-α/Anti-TNF-α oligo/pre-miR-124/anti-miR-124-3p and apoptotic death of HepG2 cells was quantified in FACS acquiring AnnexinV-FITC⁺/AnnexinV-FITC⁺PI⁺ cells. Apoptotic cell death was noted higher in presence of TNF-α compared to the mock while it was reduced upon treatment with anti-TNF-α/miR-124 and anti-miR-124-3p treatment showed opposite data ( Figure 4A ). This experiment was also repeated in HepG2 cells directly transfected with the cDNA of TNF-α/pIL1β-CC(low IL1β)/pIL1β-TT(high IL1β)/pre-miR-124 independently in presence of LPS and observed higher apoptosis level upon overexpression of both TNFα, and pIL1β-TT than mock/pIL1β-CC and restoration of miR-124 in the milieu also suppressed the apoptotic death ( Figure 4B ).

Apart from TNFR1 mediated cell death, Alcohol-induced endoplasmic reticulum (ER) stress is now established as an important contributor to apoptotic cell death in ALD (38). Thus, expression of major ER chaperon protein GRP78, two transmembrane signaling molecules ATF6 and IRE1 were quantified in HepG2 cells treated with the CM of TNF-α cDNA, Anti-TNF-α, pre-miR-124, anti-miR-124-3p transfected THP1 cells or HepG2 cells were directly transfected with cDNA of TNF-α, pIL1β-CC(low IL1β)/pIL1β-TT(high IL1β) and pre-miR-124. All the ER stress markers were significantly overexpressed in the presence of TNF-α and pIL1β-TT compared to the vector and pIL1β-CC which were dropped upon treatment with either anti-TNF-α or miR124 and again re-established upon anti-miR-124-3p treatment ( Figures 4C, D ).

ALD patients are also associated with a significant amount of liver fibrosis. Upon ethanol exposure cytokines and profibrogenic factor TGF-β released from KCs and hepatocytes, which activate quiescent hepatic stellate cells (HSCs) and transform into myofibroblast-like cells. Activated HSCs secrete collagen and extracellular matrix (ECM), and fibrosis progresses (39). In our in vitro setting either CM of TNF-α cDNA, Anti-TNF-α, pre-miR-124, Anti-miR-124-3p transfected THP1 cells was added to HSCs or HSCs were directly transfected with cDNA of TNF-α, pIL1β-CC (low IL1β)/pIL1β-TT (high IL1β) and pre-miR-124. The expression of pro-fibrogenic factor TGF-β1, myofibroblast markers Vimentin, α-SMA and Collagen1A1 was induced in presence of TNF-α or IL1β, which was suppressed upon anti-TNF-α and miR-124 treatment and regained upon restoration of Anti-miR-124-3p ( Figures 5A, B ).

Neutrophil infiltration in the liver is a hallmark of alcoholic hepatitis. The chemokines/cytokines such as IL8, MCP1, MIP2 etc. released by the LPS exposed KCs behave as chemotactic molecule for neutrophils to migrate to the site of inflammation (40). Thus, to mimic the milieu, HUVEC cells seeded on the collagen-coated upper well of Boyden chamber was treated with the CM of TNF-α cDNA/Anti-TNF-α/pre-miR-124/Anti-miR-124-3p transfected THP1 cells or directly transfected with cDNA of TNF-α/pIL1β-CC (low IL1β)/pIL1β-TT (high IL1β)/pre-miR-124 and freshly isolated neutrophils were added on top of it. Infiltrated neutrophils were counted in the lower chamber and found higher in TNF-α/IL1β overexpressed condition than the controls which was decreased upon treatment with anti TNF-α/miR-124 and reversed when treated with Anti-miR-124-3p ( Figure 5C ). Similar data was obtained upon direct transfection of plasmids in HUVEC cells ( Figure 5D ).

Next to understand the mechanism of function of miR-124-3p, the other target genes from TLR4 signaling axes were retrieved using Bioinformatics analysis. It is noteworthy to be mentioned that miR-124-3p was found to target various genes involved in LPS mediated inflammatory signaling pathways such as MyD88, TRAF6 and TRAF3; TRADD, Caspase8; IL8, MCP1, ICAM1 and TGFBR2 and PDGFRA ( Supplementary Table 3 ). The data was validated by 3’-UTR luciferase assays with WT/MT luciferase constructs of each target gene and the data has confirmed the binding of miR-124-3p to TRADD, Caspase8, PDGFRA, TGFβR2, IL8, ICAM1, and MCP1 ( Figure 6 ). Binding of miR-124-3p to the 3’UTR of other genes has been reported (31).

Cumulatively, these data suggest that alcohol sensitizes liver macrophages to secrete TNF-α and IL1β in liver microenvironment. Failure to control such situation triggers vicious activation of downstream cytokine cascade through MyD88 dependent and independent pathways leading to hepatocyte death, stellate cell activation and neutrophil infiltration. This causes progression of fibrosis and cirrhosis. Restoration of miR-124-3p in this milieu may resist the intercellular communications and withheld multiple downstream activation pathways suggesting its robust therapeutic potential for inflammatory diseases including ALD ( Figure 7 ).