Identification of Alcoholic Hepatitis-related Genes using Liver and Blood Transcriptomes

doi:10.21203/rs.3.rs-5211019/v1

Download PDF

Research Article

Identification of Alcoholic Hepatitis-related Genes using Liver and Blood Transcriptomes

https://doi.org/10.21203/rs.3.rs-5211019/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Alcoholic hepatitis (AH) is a widespread and life-threatening chronic liver condition that poses a risk of short-term mortality if not properly managed. Clinicians often encounter challenges due to insufficient knowledge about the underlying mechanisms of AH. This study employs a meta-analysis to identify the molecular mechanisms and potential cell therapy targets for AH.

Methods

We collected eight gene expression datasets, six from liver tissues and two from blood tissues, to identify AH-associated genes. Two liver datasets that had data on deaths after steroid treatment in patients with alcoholic hepatitis were also examined to uncover signatures associated with poor prognosis. Candidate genes were selected using the inverse weighted variance-based method implemented in the METAL software. We utilized prior knowledge to prioritize potential upstream genes, including a transcription factor (TF) catalog, protein-protein interaction (PPI) networks, disease-gene association databases, and summary statistics for single nucleotide polymorphisms (SNP) linked to disease and expression.

Results

Through four stepwise meta-analyses of nine gene expression datasets, we identified the robust AH liver genes. In detail, the first, second, third, and fourth steps of meta-analysis provided the liver-specific, liver-blood, and severe-mortality meta-genes linked to AH condition, respectively. Multiple lines of evidence (TF, PPI, and SNP databases) were used to identify 29 AH-related upstream genes. Among the candidates, 14 genes were replicated in the severe acute AH mouse model.

Conclusions

This study presented the candidate upstream AH genes, providing a foundation for developing AH therapeutic targets.

Alcoholic hepatitis (AH) is a life-threatening condition characterized by a necroinflammatory process leading to cirrhosis and liver failure induced by alcohol use disorder [1]. In high-risk patients, the 90-day mortality rate can reach up to 50% (as observed in the STOPHA trial) [2]. Therefore, urgent treatment decisions, such as steroids or liver transplantation, are often necessary. For decades, steroid treatment has been the primary treatment for severe AH, demonstrating improved 28-day survival in carefully indicated patients [3]. However, steroid therapy was related to a high incidence of infection and adverse events [4]. Moreover, patients not eligible for steroid therapy have limited alternatives, primarily liver transplantation [4].

Chronic alcohol consumption induces intestinal dysbiosis, leading to an increase in pathogen-associated molecular patterns (PAMP) [5]. These, in turn, activate the recruitment of inflammatory cells–Kupffer cells, macrophages, and neutrophils–and the production of inflammatory cytokines through pathogen-recognition receptors like Toll-like receptors (TLRs) and nucleotide-binding oligomerization domain-like receptors (NLRs) [6]. Other pivotal mechanisms in alcoholic liver disease (ALD) include hepatocellular injury and death [7], metabolic reprogramming [8], inflammation [9], oxidative stress [10], and a loss of liver regeneration [11]. Given the complex mechanisms of liver disease, current preclinical animal models struggle to replicate severe AH entirely [12].

To address the complex mechanisms of ALD and vulnerability to its development, genome-wide association studies (GWAS) have identified critical single nucleotides (SNP) associated with alcohol-related liver cirrhosis, such as rs738409 (PNPLA3), rs10401969 (SUGP1), rs58542926 (TM6SF2), and rs626283 (MBOAT7) [13]. However, GWAS faces limitations due to the missing heritability issue. SNPs provide only partial information about complex phenotypes and the topological challenge of many SNPs residing in non-coding regions [14]. To unbiasedly identify signatures associated with AH, liver gene expression profiling identified approximately 200 differentially expressed genes (DEGs), which were enriched in several biological pathways, including "cytokine-cytokine receptor interaction"[15]. A recent integrative study of the hepatic transcriptome and metabolome pinpointed extensive dysregulation of glucose metabolism, explicitly proposing the hexokinase domain containing 1 (HKDC1) as an essential gene related to AH [16].

Research involving genetic data from human tissues or animal experiments demands substantial effort and cost for sample collection. Individual studies were rarely able to attain the necessary sample size for generalized results, especially when faced with potential incongruent outcomes [12]. Meta-analyses are commonly employed to mitigate the lack of replication in small, heterogeneous datasets and to derive generalized findings for targeted phenotypes, particularly in GWAS studies [13]. While meta-analyses are prevalent in the GWAS domain, they are less frequently applied in transcriptomic analyses, especially in gene expression datasets profiling patients with ALD.

This study aims to identify AH-related genes and cell therapy targets based on a four-stage consecutive meta-analysis. First, a multidisciplinary team, including hepatologists, data scientists, and database administrators, conducted literature and database reviews and collected liver and blood AH-related transcriptomic datasets (Fig. 1). Using a meta-analysis approach, we identified genes with convergent patterns of change for the three lists summarizing AH-associated transcriptome changes in liver tissues as AH liver meta genes. In the second and third phases of meta-analysis, AH blood and prognosis-related datasets were pooled to discern liver-blood AH and mortality-severe meta genes, respectively (Fig. 1). In the final step, various lines of evidence, including the transcription factor (TF) database [14], disease-gene network [15], ALD-related SNPs [16], liver cis-expression quantitative trait loci (eQTL) [17], and protein-protein interaction (PPI) network [18], were utilized to provide potential upstream AH-related genes (Fig. 1).

Collection of AH-related transcriptome datasets

To obtain the meta-analyzed AH-related signatures, we collected six liver AH transcriptome datasets (GSE28619, GSE103580, GSE143318, GSE142530, GSE167308, and GSE155907, Figure S1 – S6), along with two blood AH dataset (GSE135285 and GSE171809). Additionally, we compiled two liver gene expression datasets (GSE94397 and GSE94399) to identify pooled signatures associated with the poor prognosis of AH.

To systematically compare the compiled gene expression datasets, we unified the transcript IDs in each dataset using Entrez Gene [19]. Chip-specific transcripts, probes, and probe-set IDs were contained in the microarray datasets, while Ensembl IDs were implemented for all genes in the RNA-seq datasets. We systematically annotated them into gene symbol approved by the HUGO Gene Nomenclature Committee [20] using the org.Hs.eg.db, org.Mm.eg.db, and Orthology.eg.db packages to perform the meta-analysis.

Differential expression analysis

Differential expression (DE) analyses of microarrays and RNA-seq were performed to compare specific phenotypes (e.g., AH and death) with their matched controls. The limma and DESeq2 methods were employed for microarray and RNA-seq data, respectively. Both DE tools provided gene-specific fold changes (FC) and matched p-values for comparing disease and control samples. Genes with a false discovery rate (FDR)-adjusted p-value of less than 0.5 were identified as differentially expressed genes (DEGs) between the two conditions.

A hypergeometric test was applied to identify enriched biological pathways in the candidate gene sets. Enrichment analyses were carried out using the Gene Ontology (GO) [21] and Kyoto Encyclopedia of Genes and Genomes (KEGG) [22] databases. Biological pathways were considered enriched when they met the criteria of a 0.5 FDR-adjusted p-value determined by the hypergeometric test.

Meta-analysis

We conducted a meta-analysis to identify biomarkers commonly associated with diseases in the collected candidate gene expression datasets. The meta-analysis utilized a fixed-effect inverse-variance-weighted (IVW) method adopted in METAL software [23]. The IVW method requires an effect similar to the FC between the two conditions and a matched standard error. We selected gene-specific FC between the two conditions for each transcriptomic dataset and matched the standard errors obtained from summary statistics calculated using limma or DESeq2 in the previous step. The significance level for associations calculated through meta-analysis was determined by FDR adjustment for multiple comparison test.

Bayesian approach for Identifying Upstream Biomarkers

Recent studies implemented multiple lines of evidence as prior knowledge to prioritize the candidate upstream genes [24–29]. Hägg et al.[24] curated gene sets involved in transcription activities such as transcription activator activity, transcription coactivator activity, and TF binding from the GO database [21] to identify upstream genes related to the coronary artery. Lee and Lee [25] used a TF-related gene set obtained from a TF database to identify Alzheimer’s disease (AD)-related genes. Recently, Lambert et al.[14] proposed a TF catalog integrating prominent TF databases, and several studies [26, 27] used it to narrow down the potential gene sets correlated to disease and comorbidity. Motivated by these studies, we implemented the TF catalog manually updated by Lambert et al.[14] to identify the upstream genes related to AH (Fig. 1).

We implemented the DigSee database by curating the relationships of approximately 4500 disease types with about 13000 genes by integrating text mining and machine learning methods [15, 30]. Using a “Liver Diseases, Alcoholic” query, approximately 500 ALD-related genes were compiled from the DigSee.

A GWAS was conducted on alcohol-related cirrhosis in separate German and UK cohorts (712 cases and 1,426 controls) [16]. Full summary statistics of the alcohol-related cirrhosis GWAS were downloaded from the supplementary website (http://gengastro.med.tu-dresden.de/suppl/alc_cirrhosis/) [16]. Among the 6,502,449 SNPs, approximately 5500 variants with uncorrected p-values < 0.001 for the association between genetic variants and alcohol-related cirrhosis were selected. The 5500 SNPs were assigned to their corresponding genes using the Gsnpense function in the gprofiler2 package [31]. Among the candidate AH genes, those with evidence from GWAS were selected as potential biomarkers of AH (Fig. 1).

Liver cis-eQTL data were obtained from the Genotype-Tissue Expression (GTEx) project [17]. In detail, full summary statistics of the GTEx liver eQTL were obtained with an access ID (study ID: QTS000015; dataset ID: QTD000266) from the eQTL catalog database [32]. Cis-associations between gene expression and variant types with an uncorrected p-value < 10^–5 were selected, yielding 317,324 gene-SNP pairs that accounted for 2829 genes (annotated by symbols assigned by HGNC). Among the candidates the meta-analysis chose, genes showing evidence of liver cis-eQTL were designated candidate upstream genes (Fig. 1).

A PPI network was obtained from the STRING database, which collected the interactome from various sources, including automated text mining of scientific and medical literature, computational predictions of co-expression and co-occurrence across genomes, PPI experiment databases, and known biological pathways [18]. STRING consists of approximately 12 million edges of 40,000 proteins (based on the Ensembl Protein (ESPN)). The PPI network introduced by the STRING database included protein-protein pairs and their matched scores. The distribution of PPI scores did not follow a Gaussian distribution; therefore, PPIs with the top 90 percent of the interactome scores were selected. Among significant genes in meta-analysis, genes with ≥ 200 edges were chosen as the candidate upstream genes (Fig. 1).

Severe acute AH mouse model

Mice (C57BL/6, male, seven weeks) were randomly assigned to AH and control groups. For the AH group, mice were administered oral ethanol (5 g/kg/day) and intraperitoneal thioacetic acid (TAA, 200mg/kg, twice week/4 weeks), and the body weight was measured every week, and biochemistry analyses were performed by collecting blood before sacrifice. For the control group, mice were injected with normal saline in peritoneal space.

After four weeks of the mouse experiments, mice were anesthetized with isoflurane inhalation. Liver were extracted and fixed in 4% paraformaldehyde, imbedded in paraffin, followed by dehydration in graded alcohol. Tissue slices of 5 µm thickness were prepared and stained with hematoxylin-eosin (Sigma-Aldrich, St. Louis, MO, USA).

RNA isolation and real-time polymerase chain reaction

Total RNA was extracted from frozen whole livers using TRIzol reagent (Invitrogen, Carlsbad, CA) or Qiagen mini columns (Qiagen Inc. Valencia, CA) according to the manufacturer’s protocol. RNA concentrations were quantified by spectrophotometry. cDNA was synthesized form total RNA (1ug) by using the RT Premix kit (TOYOBO, Japan). Real-time PCR was performed with the QuantStudioTM6 Flex instrument (Applied Biosystems) using SYBR Green PCR master mix (Applied Biosystems) and the respective primer pairs for each gene. Data were analyzed by using QuantStudio 6 and 7 Flex Software (Applied Biosystems). The cycle threshold (Ct) values of the target genes were normalized to those of endogenous control gene (beta-actin or GAPDH). Relative expression of mRNAs for various genes is presented as fold change using the 2^ (-ΔΔCT) method.

Meta-analysis of the liver AH gene expression datasets

The hierarchical clustering of genes with the top 80 variances across liver samples (GSE28619) led to a distinct separation between AH and control samples (Figure S1). PCA of the varied genes of liver tissues demonstrated a clear classification between AH and control (Figure S1). In other liver datasets (GSE103580, GSE143318, GSE142530, GSE167308, and GSE155907), hierarchical clustering and PCA revealed a noticeable classification between AH and control groups (Figures S2–6). Four liver datasets (GSE103580, GSE143318, GSE142530, and GSE167308) included alcoholic liver cirrhosis (ALC) cases, and most sets, except for GSE103580, provided the convergent findings that the transcriptomic signatures of ALC were placed midpoint between those of control and AH (Figures S2 – S5).

Six lists of DEGs between the AH and control groups were curated based on an FDR-adjusted p-value threshold of 0.05 from the seven liver transcriptomic datasets. Only two genes (AKR1B10 and LTBP2) were AH-related genes in all six liver AH datasets (Table S1). Two hundred eighty genes were contained in five liver AH-DEG lists (Table S1), among which ADRB2, AQP1, CFTR, EPCAM, ESR1, HPGD, IL18, IL6R, MDK, MMP2, SPP1, SRC, THBS2, VCAN, VIM, and VMF are estimated to interact with drugs [33].

The global transcriptomic signatures related to AH in GSE167308 were placed in the hub position among six liver gene expression datasets (Fig. 2). Six liver gene expression datasets were meta-analyzed to identify generalized liver genes associated with AH status. Specifically, the summary statistics from the six gene expression datasets, calculated using the limma and DESeq2 algorithms [34, 35], were integrated using the IVW method in METAL [23]. This analysis identified approximately 3,400 liver genes related to AH, based on an FDR-adjusted p-value < 0.05, and these genes were annotated as liver AH meta-genes (Fig. 2). Additionally, 91 genes that were not identified as DEGs in any of the six liver datasets were discovered through the meta-analysis (Figure S7). AKR1B10, DTNA, SPP1, GOLM1, CXCL6, CTNNA3, KCNN2, KCNN3, ST3GAL6, and SMOC1 were included in the top 50 liver AH meta genes (Fig. 2). The top 200 liver meta-genes showed slight biases to upregulated expression in AH groups (Figure S8).

The liver AH meta-genes were found to be associated with oxidoreductase activity, arachidonic acid monooxygenase activity, NADPH activity, alpha-amino acid catabolic process, xenobiotic catabolic process, taurine metabolic process, Retinol metabolism, beta-Alanine metabolism, and primary bile acid biosynthesis (Fig. 2). The following pathways mainly included up-regulated genes in AH status: platelet-derived growth factor binding, basement membrane collagen trimer, and complex of collagen trimers (Fig. 2).

Identification of the blood and mortality-severe AH meta-genes

Global transcriptomic signatures related to AH of two blood datasets (GSE135285 and GSE171809) were curated using DESeq2 [35]. Both blood gene expression datasets contained dominant proportions of AH-DEG upregulated in AH status (Figure S9). We meta-analyzed the AH liver meta-genes with two blood AH-related signatures. As a result, approximately 2,560 DEGs with an FDR-adjusted p-value < 0.05 were identified by IVW-based meta-analysis and categorized as liver-blood AH meta-genes (Fig. 3). 1,498 genes were not observed in individual liver and blood datasets within the liver-blood meta-genes but were newly identified in the meta-analysis (Fig. 3). The top 50 blood genes having high absolute meta-analyzed z-values were composed predominantly of up-regulated genes in AH status (Fig. 3). The 2,560 blood AH meta-genes were estimated to involve the following biological pathways: ATPase-coupled intramembrane lipid transporter activity, phosphatidylcholine floppase activity, GTPase binding, and IgG immunoglobulin complex (Fig. 3). Genes included in fibroblast proliferation, subtelomeric heterochromatin formation, small-cell lung cancer, and focal adhesion were upregulated in liver and blood AH cases (Fig. 3).

For the two liver AH prognosis-related gene expression datasets (GSE94397 and GSE94399), we curated two lists of FC values with matched standard errors for all genes in the form of summary statistics calculated by limma (Figure S10). Transcriptomic alteration in the AH death group exhibited weak signals (Figure S10). Based on uncorrected p-values less than 0.001, only 20 and 18 genes were identified as the AH-DEG in GSE94397 and GSE94399, respectively. According to a nominal p-value less than 0.01, the GSE94397 and GSE94399 provided 195 and 203 AH-DEGs, respectively (Figure S11). Using IVW-based meta-analysis, we integrated three lists of summary statistics for all genes, including liver-blood AH meta-genes, and two liver summary statistics encompassing AH-related death signatures. The IVW method identified approximately 2,500 genes related to AH and death due to AH after steroid treatment (Fig. 4). We defined these about 2,500 genes as mortality-severe AH meta-genes, among which the top 100 genes with high absolute z-values included predominantly down-regulated genes (Fig. 4). The 2,500 AH death meta-genes were estimated to contribute to several pathways, such as taurine metabolic processes, peroxisome, and high-density lipoprotein particles. Most expected biological pathways mainly included down-regulated genes.

External Biological Evidence for Identifying Upstream AH-related Genes

Integrative bio-digital studies have attempted to prioritize candidate crucial genes and biomarkers by integrating and tabulating multiple lines of evidence. Niculescu et al.[36] introduced Convergent Functional Genomics (CFG), a method to narrow down biologically relevant features in a Bayesian fashion. As one part of the Bayes approach, Hägg et al.[24] identified the LDB2 gene as the CVD regulator using TF factor information obtained from the GO database. Among the about 2,500 AH meta-genes, 350 markers were TF-related genes confirmed by the TF catalog [14] and were scored as 1 (Figure S12 and Table S2).

A study exploring Alzheimer's disease-related genes assigned a high CFG score when a gene is hit by GWAS, eQTL, and PPI evidence [29]. Several Bayes-based studies implemented TF, disease-gene networks, GWAS, eQTL, and PPI databases to identify regulators or upstream genes [25–29]. Taking advantage of these frameworks, prior knowledge, including the TF catalog [14], disease-gene relationship database [15, 30], full summary statistics of GWAS [16], liver eQTL [17], and protein interactome [18] was implemented to uncover the upstream AH-related biomarkers. As a result, the Bayes approach narrowed down 29 genes among the about 2,500 AH mortality-severe meta-genes (SREBF2, TGFB1, GSTP1, LYN, MERTK, TLR2, HDAC9, ZNF217, ZFP90, MAPK3, PDE4D, SEMA5A, NR1I2, KRAS, CBS, PCBP2, PPARA, NFE2L2, AGT, MAPK8, ESR1, MET, PPARGC1A, LPIN1, SELE, DNAJB1, LIPC, NR3C1, AFP) had three or more of external evidence (Fig. 5), which were selected as the candidate upstream AH genes.

Severe acute AH mouse model

A severe acute AH (SAAH) mouse model was established by administering oral ethanol and intraperitoneal TAA for 4weeks (Figures S13 – S14). The SAAH model exhibited the following phenotypes: weight loss, enlarged gallbladder, high liver profiles, and lobular necroinflammation (Figures S14 – S15). Targeted mRNA measurements of liver tissues of the SAAH model were performed on mouse liver tissue for 29 candidate AH genes. As a result, the downregulation of hepatic expression of SREBF2 was replicated in the severe acute AH mouse model (Fig. 6). Collectively, the downregulation of SREBF2 had been demonstrated in three rounds of meta-analysis (liver, blood, and death) and in-vivo experiments. Moreover, multiple lines of external evidence, including TF-related genes [14], ALD-related genes [15], GWAS [16], and high connectivities in protein interactome [18], suggested SREBF2 as the upstream genes.

In the same way, the following 13 genes showed the convergent findings between AI and in vivo experiments: TGFB1, LYN, TLR2, HDAC9, CBS, PCBP2, PPARA, AGT, MET, LPIN1, DNAJB1, LIPC, and NR3C1 (Fig. 6). Among 14 final AH genes, TGFB1, LYN, TLR2, and HDAC9 were identified as the robust up-regulated genes that are the candidate targets for inhibitors.

We identified eight lists of meta-genes related to AH, designated as liver AH (Fig. 2) and liver blood AH (Fig. 3). Subsequently, gene expression datasets associated with mortality after steroid therapy were utilized to generate candidate lists of genes targeted by intensive treatments, such as immune modulators and mesenchymal stem cells [37], annotated as mortality-severe meta-genes (Fig. 4). Employing validated and globally recognized databases, such as the TF catalog [14], disease-gene network [15, 30], and protein interactome [18], we selected candidate upstream genes crucial in AH that could potentially serve as therapeutic targets (Fig. 5).

The analysis included some genes not previously characterized in the context of ALD. Transforming growth factor-β (TGF-β) is a critical regulator in the progression of chronic liver disease, playing a role at every stage of the disease. When TGF-β is activated, it promotes hepatocyte destruction and triggers the activation of hepatic stellate cells, leading to fibroblast proliferation and extracellular matrix deposition. This cascade significantly contributes to liver fibrosis. [38] Our previous studies have demonstrated the pivotal role of TGF-β in a cirrhosis animal model, highlighting its potential as a therapeutic target for chronic liver disease. [39]

Lyn kinase is a tyrosine receptor for liver sinusoidal endothelial cells (LSECs), specialized fenestrated endothelial cells that maintain liver homeostasis by regulating intrahepatic vascular tone, immune cell function, and the quiescence of hepatic stellate cells (HSCs). [40] In a CCL4-induced animal model, Lyn expression was significantly increased, and this upregulation was linked to fibrosis through HSC activation. When TGF-β was administered, overexpressed Lyn kinase led to an increase in collagen and α-SMA, suggesting Lyn's involvement in TGF-β-mediated hepatic fibrosis. [41] Additionally, Lyn kinase inhibitor caused the apoptosis of activated HSCs. Therefore, it is presumed that Lyn kinase is involved in the progression of hepatic fibrosis through TGF-β signaling and HSC activation. Many tyrosine kinase inhibitors, which target enzymes like Lyn kinase, are already used in cancer therapies.

TLR2 is expressed in various liver cells, including Kupffer cells, hepatic stellate cells (HSCs), hepatocytes, and liver sinusoidal endothelial cells (LSECs) [42] Damaged hepatocytes released DAMPs (e.g., heat shock proteins, HMGB1, and other cellular debris). Then, TLR2 recognizes DAMPs and microbial components (PAMPs), such as lipoproteins and lipoteichoic acid, from bacterial cell walls, initiating the immune response[43, 44]. Upon activation by TLR2 ligands, the Kupffer cell secretes pro-inflammatory cytokines. These cytokines further activate hepatic stellate cells (HSCs) [45] TLR2 activation also induces the release of fibrogenic factors like TGF-β [46]. Given its role in liver fibrosis, TLR2 serves as a potential therapeutic target. Inhibiting TLR2 could help reduce inflammation and fibrogenesis in the liver, potentially slowing or reversing fibrosis progression.

Histone deacetylases (HDACs) regulate the histone acetylation code dynamically and reversibly. Changes in mRNA levels of specific HDACs during hepatic fibrogenesis appear complex. A recent study demonstrated that CCl₄-induced liver fibrosis promoted the upregulation of several HDAC, while fibrosis reversal was accompanied by the downregulation of the expression of specific HDACs [47]. In contrast, a binge alcohol model showed the downregulation of HDAC9 expression. [48] These discrepancies between studies could be model-dependent, highlighting the need for further research to establish global HDAC expression patterns during hepatic fibrogenesis. Notably, HDAC9 expression is also markedly elevated in human livers, including primary biliary cirrhosis (PBC), alcoholic cirrhosis, and NASH [49]. Additionally, HDAC inhibitors are now widely addressed as therapeutic drugs for human diseases and are considered safe. Our findings support HDAC9 as a promising therapeutic target for liver fibrosis.

This study employed a multidisciplinary approach to meticulously screen candidate genetic data, followed by their integration and analysis. We aimed to minimize potential biases within each cohort by utilizing four datasets based on statistical methods. A notable strength of our research lies in the combination of blood-derived gene data, a strategy employed to overcome the limitations associated with liver tissue analysis. Previous studies have suggested the utility of genomic indicators from blood as biomarkers for predicting liver disease [50]. Consistent with a prior study, blood-based diagnostic gene biomarkers derived from peripheral blood mononuclear cells (PBMC) demonstrated a 90% overall accuracy in differentiating AH from alcoholic cirrhosis [51]. Consequently, we incorporated this hypothesis into our study, leveraging blood-derived samples to reflect liver disease and its state. Furthermore, genes identified through this analysis underwent cross-analysis with datasets related to prognosis, enabling the selection of genes capable of predicting AH prognosis or monitoring treatment response in clinical settings. This allows for the establishment of a model that predicts the response to AH treatment or provides a targeted mechanism for its management.

In conclusion, our meta-analysis aimed to identify the robust AH genes. Additionally, by incorporating multiple lines of external evidence as prior knowledge, we estimated the upstream markers of AH that could serve as potential therapeutic targets.

Conflict of Interest Statement

All authors have no conflicts of interest to declare.

Ayares G et al (2022) Current Medical Treatment for Alcohol-Associated Liver Disease. J Clin Experimental Hepatol 12(5):1333–1348
Forrest E et al (2013) Steroids or pentoxifylline for alcoholic hepatitis (STOPAH): study protocol for a randomised controlled trial. Trials 14:262
European Association for the Study (2012) of, L., EASL clinical practical guidelines: management of alcoholic liver disease. J Hepatol 57(2):399–420
Thursz MR et al (2015) Prednisolone or pentoxifylline for alcoholic hepatitis. N Engl J Med 372(17):1619–1628
Bataller R, Arab JP, Shah VH (2022) Alcohol-Associated Hepat N Engl J Med 387(26):2436–2448
Saha B et al (2019) Biomarkers of Macrophage Activation and Immune Danger Signals Predict Clinical Outcomes in Alcoholic Hepatitis. Hepatology 70(4):1134–1149
Miyata T, Nagy LE (2020) Programmed cell death in alcohol-associated liver disease. Clin Mol Hepatol 26(4):618–625
Sun R et al (2022) Alcohol-driven metabolic reprogramming promotes development of RORγt-deficient thymic lymphoma. Oncogene 41(16):2287–2302
Wang HJ et al (2012) Inflammation in alcoholic liver disease. Annu Rev Nutr 32:343–368
Tsermpini EE, Plemenitaš Ilješ A, Dolžan V (2022) Alcohol-Induced Oxidative Stress and the Role of Antioxidants in Alcohol Use Disorder: A Systematic Review. Antioxidants (Basel), 11(7)
Lv Y, So KF, Xiao J (2020) Liver regeneration and alcoholic liver disease. Ann Transl Med 8(8):567
Lewis M et al (2022) The puzzling relationship between multi-laboratory replications and meta-analyses of the published literature. Royal Soc Open Sci 9(2):211499
Zhou W et al (2022) Global Biobank Meta-analysis Initiative: Powering genetic discovery across human disease. Cell Genom 2(10):100192
Lambert SA et al (2018) The Human Transcription Factors. Cell 172(4):650–665
Kim J, Kim JJ, Lee H (2017) An analysis of disease-gene relationship from Medline abstracts by DigSee. Sci Rep 7:40154
Buch S et al (2015) A genome-wide association study confirms PNPLA3 and identifies TM6SF2 and MBOAT7 as risk loci for alcohol-related cirrhosis. Nat Genet 47(12):1443–1448
Lonsdale J et al (2013) The Genotype-Tissue Expression (GTEx) project. Nat Genet 45(6):580–585
Szklarczyk D et al (2023) The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res 51(D1):D638–d646
Maglott D et al (2007) Entrez Gene: gene-centered information at NCBI. Nucleic Acids Res 35(Database issue):D26–31
Povey S et al (2001) The HUGO Gene Nomenclature Committee (HGNC). Hum Genet 109(6):678–680
The Gene Ontology Resource: 20 years and still GOing strong. Nucleic Acids Res, (2019) 47(D1): p. D330–d338
Kanehisa M, Goto S (2000) KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28(1):27–30
Willer CJ, Li Y, Abecasis GR (2010) METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics 26(17):2190–2191
Hägg S et al (2009) Multi-organ expression profiling uncovers a gene module in coronary artery disease involving transendothelial migration of leukocytes and LIM domain binding 2: the Stockholm Atherosclerosis Gene Expression (STAGE) study. PLoS Genet 5(12):e1000754
Lee T, Lee H (2020) Prediction of Alzheimer’s disease using blood gene expression data. Sci Rep 10(1):3485
Lee T, Lee H (2021) The Alzheimer's Disease Neuroimaging, Identification of Disease-Related Genes That Are Common between Alzheimer's and Cardiovascular Disease Using Blood Genome-Wide Transcriptome Analysis. Biomedicines 9(11):1525
Koplev S et al (2022) A mechanistic framework for cardiometabolic and coronary artery diseases. Nat Cardiovasc Res 1(1):85–100
Zhang B et al (2013) Integrated systems approach identifies genetic nodes and networks in late-onset Alzheimer's disease. Cell 153(3):707–720
Xu M et al (2018) A systematic integrated analysis of brain expression profiles reveals YAP1 and other prioritized hub genes as important upstream regulators in Alzheimer's disease. Alzheimers Dement 14(2):215–229
Kim J et al (2013) DigSee: Disease gene search engine with evidence sentences (version cancer). Nucleic Acids Res, 41(Web Server issue): p. W510–W517
Kolberg L et al (2020) gprofiler2–an R package for gene list functional enrichment analysis and namespace conversion toolset g: Profiler. F1000Research, 9
Kerimov N et al (2021) A compendium of uniformly processed human gene expression and splicing quantitative trait loci. Nat Genet 53(9):1290–1299
Cannon M et al (2023) DGIdb 5.0: rebuilding the drug–gene interaction database for precision medicine and drug discovery platforms. Nucleic Acids Res 52(D1):D1227–D1235
Smyth GK (2005) Limma: linear models for microarray data, in Bioinformatics and computational biology solutions using R and Bioconductor. Springer, pp 397–420
Love MI, Huber W, Anders S (2014) Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15(12):550
Niculescu AB 3 et al (2000) Identifying a series of candidate genes for mania and psychosis: a convergent functional genomics approach. Physiol Genomics 4(1):83–91
Lee T et al (2023) Comparative Analysis of Biological Signatures between Freshly Preserved and Cryo-Preserved Bone Marrow Mesenchymal Stem Cells. Cells 12(19):2355
Dooley S, Dijke P (2012) TGF-beta in progression of liver disease. Cell Tissue Res, 347(1): pp. 245 – 56
Jang YO et al (2018) Synergistic effects of simvastatin and bone marrow-derived mesenchymal stem cells on hepatic fibrosis. Biochem Biophys Res Commun 497(1):264–271
de Jesus AA et al (2023) Constitutively active Lyn kinase causes a cutaneous small vessel vasculitis and liver fibrosis syndrome. Nat Commun 14(1):1502
Li Y, Xiong L, Gong J (2017) Lyn kinase enhanced hepatic fibrosis by modulating the activation of hepatic stellate cells. Am J Transl Res 9(6):2865–2877
Getachew A et al (2021) Toll-like receptor 2 signaling in liver pathophysiology. Life Sci 284:119941
Takeuchi O, Akira S (2010) Pattern recognition receptors and inflammation. Cell 140(6):805–820
Jagavelu K et al (2010) Endothelial cell toll-like receptor 4 regulates fibrosis-associated angiogenesis in the liver. Hepatology 52(2):590–601
Friedman SL (2008) Mechanisms of hepatic fibrogenesis. Gastroenterology 134(6):1655–1669
Aoyama T et al (2010) CX3CL1-CX3CR1 interaction prevents carbon tetrachloride-induced liver inflammation and fibrosis in mice. Hepatology 52(4):1390–1400
Li X et al (2016) Role of histone deacetylases(HDACs) in progression and reversal of liver fibrosis. Toxicol Appl Pharmacol 306:58–68
Kirpich I et al (2012) Binge alcohol-induced microvesicular liver steatosis and injury are associated with down-regulation of hepatic Hdac 1, 7, 9, 10, 11 and up-regulation of Hdac 3. Alcohol Clin Exp Res, 36(9): pp. 1578-86
Yang Y et al (2017) Histone deacetylase 9 plays a role in the antifibrogenic effect of astaxanthin in hepatic stellate cells. J Nutr Biochem 40:172–177
Huang J et al (2010) Genomic indicators in the blood predict drug-induced liver injury. Pharmacogenomics J 10(4):267–277
Listopad S et al (2022) Differentiating between liver diseases by applying multiclass machine learning approaches to transcriptomics of liver tissue or blood-based samples. JHEP Rep 4(10):100560

The authors declare no competing interests.

SUP240929.pdf
Supplemenatry Materials for Identification of Alcoholic Hepatitis-related Genes using Liver and Blood Transcriptomes

Download PDF

Version 1

posted

You are reading this latest preprint version

Identification of Alcoholic Hepatitis-related Genes using Liver and Blood Transcriptomes

Status:

Version 1

Abstract

Figures

Introduction

Method

Collection of AH-related transcriptome datasets

Differential expression analysis

Meta-analysis

Bayesian approach for Identifying Upstream Biomarkers

Severe acute AH mouse model

RNA isolation and real-time polymerase chain reaction

Result

Meta-analysis of the liver AH gene expression datasets

Identification of the blood and mortality-severe AH meta-genes

External Biological Evidence for Identifying Upstream AH-related Genes

Severe acute AH mouse model

Discussion

Declarations

Conflict of Interest Statement

References

Additional Declarations

Supplementary Files

Status:

Version 1