{"paper_id":"d65de501-eeb9-4cc4-a7b5-09df8ac41e0d","body_text":"long noncoding RNAs\nhepatic nuclear factor 4 alpha, opposite strand\nhepatocyte nuclear factor 4 alpha\nPPARγ coactivator-1α\nreactive oxygen species; I/R, ischemia/reperfusion\nanoxia/reoxygenation\nalanine aminotransferase\naspartate aminotransferase\nenzyme-linked immunosorbent assay\nterminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling\nmalondialdehyde\n4-hydroxynonenal\nsuperoxide dismutase\ncatalase\nglutathione peroxidase\nlactate dehydrogenase\n\nHepatic ischemia/reperfusion injury (HIRI) is a common pathological process that occurs in several clinical scenarios, such as complex liver resection, liver transplantation, and hemorrhagic shock. During this process, the initial ischemic injury causes direct hepatocyte damage, and subsequent blood flow reflux further aggravates liver dysfunction and injury due to the propagation of reactive oxygen species (ROS), macrophage activation and inflammatory cytokines, which trigger cell death [ 1 , 2 ]. However, the underlying molecular mechanisms of ischemia/reperfusion (I/R) injury remain largely unknown.\nLong noncoding RNAs (lncRNAs) are defined as single-stranded RNA molecules spanning more than 200 nucleotides that are involved in multilevel gene expression regulation, including epigenetic modification, and transcriptional and posttranscriptional progression [ 3 ]. According to the proximity to protein coding genes in the genome, lncRNAs are generally placed into five categories: sense, antisense, bidirectional, intronic, and intergenic lncRNAs [ 4 ]. Currently, several studies have highlighted the significant roles of lncRNAs in the pathogenesis of liver disease. For instance, lncRNA  HULC  is upregulated in hepatocellular carcinoma and enhances hepatocarcinogenesis by promoting the phosphorylation of YB-1 via the ERK pathway [ 5 ]; lncRNA  ANRIL  alleviates liver fibrosis and hepatic stellate cell (HSC) activation via the AMPK pathway [ 6 ]; and lncRNA  CCAT1  promotes nonalcoholic fatty liver disease (NAFLD) by increasing LXRα transcription [ 7 ]. Nevertheless, in the case of hepatic I/R injury, little is known about lncRNAs in hepatic I/R injury. Thus, a deeper understanding of the molecular mechanisms underlying the pathogenic process of hepatic I/R is required to uncover potential lncRNA-targets for developing promising therapeutic strategies.\nFurthermore, we have identified a novel lncRNA hepatic nuclear factor 4 alpha, opposite strand ( Hnf4αos ), a natural antisense transcript (NAT) of hepatocyte nuclear factor 4 alpha (Hnf4α), which was aberrantly upregulated in mouse I/R models. Although  Hnf4αos  has been reported, little information is available for regarding its molecular function [ 8 , 9 ]. PPARγ coactivator 1 alpha (PGC1α) is well known as a metabolic regulator in the physiological process of oxidative phosphorylation (OXPHOS), the tricarboxylic acid (TCA) cycle and ROS metabolism [ [10] ,  [11] ,  [12] ]. Intriguingly, our previous studies have demonstrated that PGC1α is an important regulator of ROS metabolism that reduces cell death, ameliorates the sterile inflammatory response and alleviates oxidative stress-induced liver damage during hepatic I/R insult [ 13 ]. Moreover, several lines of evidence, including data from bioinformatic analysis and determination of oxidative stress levels, suggest a close link between the lncRNA  Hnf4αos  and PGC1α. Thus, we further investigated the effects of  Hnf4αos  on I/R progression and the underlying mechanisms between  Hnf4αos  and PGC1α.\n\nHuman liver samples were obtained from subjects who underwent partial hepatectomy due to hepatic hemangioma. All procedures involving human samples were approved by the Ethics Committee of the First Affiliated Hospital of Harbin Medical University and patient informed consent was obtained. We listed the detailed clinical information of the hemangioma patients in  Supplementary Table S3 .\nMale C57BL/6 mice, hepatocyte-specific  Hnf4αos  knockout ( Hnf4αos -KO) mice and wild-type (WT) mice (8 weeks old) were housed in specific pathogen-free (SPF) conditions and raised following institutional guidelines for animal care.  Hnf4αos -KO mice were obtained by CRISPR/Cas9 methods as described previously [ 14 ].  Hnf4αos -KO mice were generated by crossing  Hnf4αos -floxed mice with Albumin-Cre mice (Jackson Laboratory. Bar Harbor, ME, USA) on the C57BL background. The donor vector containing the fourth exon of the  Hnf4αos  gene was floxed by two loxP sites. All animal experiments were performed in accordance with the standard protocols of the Committee on the Use of Live Animals in Teaching and Research of Harbin Medical University, Harbin, China.\nThe procedures for partial hepatic ischemia have been described previously [ 15 ]. Mice were housed in a specific pathogen-free and temperature-controlled environment with a 12-h light/dark cycle. Briefly, the mice were anesthetized with pentobarbital sodium (50 mg/kg), and a midline laparotomy was performed. An atraumatic clip was placed across the left lateral and median lobes of the liver (∼70%). After 75 min of partial hepatic ischemia, the clip was removed for initial reperfusion. Sham control mice underwent the same operation without vascular clamping.\nCellular anoxic conditions were established and maintained in a modular incubator chamber (Biospherix, Lacona, NY, USA) by continuous gas flow with a 1% O 2 , 5% CO 2  and 94% N 2  gas mixture. After incubation under hypoxia for 6 h, the cells were incubated under normoxic conditions with 95% air and 5% CO 2  for the indicated times (0, 3, 6, 12, 24h). The medium and cells were collected for further analysis.\nMouse hepatocytes were isolated by a modified in situ collagenase perfusion technique as previously described [ 15 ]. Hepatocyte purity and viability typically exceeded 99 and 95%, respectively. Primary hepatocytes and L02 cell lines (Type Culture Collection of the Chinese Academy of Science) were cultured in DMEM supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin in a 5% CO 2 /water-saturated incubator at 37 °C.\nParaffin-embedded tissue sections were used for immunofluorescence as described previously [ 16 ]. The liver sections were incubated with primary antibody against Ly6G (Cell Signaling Technology) (1:500) (31469), and the slides were incubated with corresponding fluorescence-labeled secondary antibody (ThermoFisher) (1; 1000) (A32744) for further staining.\nCellular reactive oxygen species (ROS) levels were estimated as previously described [ 17 ]. For intracellular ROS levels, cells were incubated in medium containing 10 μM dihydroethidium (DHE) (Invitrogen, USA) for 30 min at 37 °C in the dark. The medium was switched to fresh medium before fluorescence detection. The relative ROS levels, which are proportional to the fluorescence intensity, were quantified using Image-Pro Plus software.\nWe predicted potential Hnf4α binding sites on the PGC1α and miR-23a promoters using the JASPAR database, and the PGC1α 3’-untranslated region (UTR) contains conserved miR-23a binding sites as reported previously [ 18 ]. We then cloned the candidate binding sites in an SV40 driven luciferase reporter plasmid. Briefly, luciferase activity was assessed using a luciferase assay kit (Promega, Madison, WI, USA). HEK-293T cells containing specific plasmids and 1 ng pRL-TK Renilla luciferase plasmid were seeded into 24-well plates. After 48 h, we used the dual luciferase reporter assay system (Promega) to measure luciferase activity according to the manufacturer's instructions.\nA ribonuclease protection assay (RPA) and quantitative RT-PCR were performed to detect the RNA-RNA duplex. Total RNA from primary hepatocytes was isolated as described previously [ 19 ]. The RNA samples were treated with DNAse Ⅰ (Sigma, 12.5 units/ml) and RNase A (QIAgen, 200 ng/ml) to remove residual DNA and single-stranded RNAs. Finally, the solutions were incubated for 40 min at 37 °C for further qRT-PCR.\nAn electrophoretic mobility shift assay (EMSA) was performed as described previously [ 12 ]. The oligonucleotides used in EMSA were as follows: Hnf4α/miR-23a wt, 5’-GATCAGCTGGCCCCTGAAAACCTTGTTTAAC-3’ and 3’-CTAGTCGACCGGGGACTTTTGGAACAAATTG-5’. Hnf4α/miR-23a mut, 5’-GATCAGCTCCCCCCTAAAAAACTTGTTTAAC-3’ and 3’-CTAGTCGAGGGGGGATTTTTTGAACAAATTG-5’.\nAll data are expressed as the mean ± SD. Significant differences between groups were determined by ANOVA, with Bonferroni correction for continuous variables and multiple groups. Two-tailed Student's  t -test was used for comparison of a normally distributed continuous variable between 2 groups. The level of significance was set at a p value less than 0.05 for all analyses.\nFurther details of the experimental materials and procedures are described in the Supplementary Files.\n\nSeveral lncRNAs were differentially expressed in the GEO data-set ( GSE15891 ) with exposure to chronic anoxia and our heatmap demonstrated the marked differentially expressed lncRNAs related to oxidative stress, inflammatory response and apoptosis pathways ( Fig. 1 A). For examining the relationships of lncRNAs and traget genes, the top-ranked lncRNAs and mRNAs correlated oxidative stress/inflammatory response/apoptosis resident on different chromosomes ( Fig. 1 B). Among the top-ranked differentially expressed lncRNAs, only  Hnf4αos  was enriched in adult mouse liver tissue ( Supplementary Table S1, 2 ). Thus,  Hnf4αos  was selected for further investigation during hepatic I/R injury. To explore the role of lncRNA  Hnf4αos  in HIRI, we first detected the expression levels of  Hnf4αos  in murine hepatic I/R and hepatocyte A/R models, and  Hnf4αos  was found to be increased after reperfusion. The human-derived lncRNA,  Hnf4α-as1 , was also found to be differentially expressed in clinical liver samples from patients who underwent partial hepatectomy ( Fig. 1 C–D,  Supplementary Fig. S1 ). Furthermore, cellular fractionation of hepatocytes followed by qRT-PCR implied that  Hnf4αos  was predominantly expressed in the nuclei of hepatocytes rather than other compartments, compared with U6 (localized in the nucleus) and 18S (localized in the cytoplasm) expression ( Fig. 1 E). Moreover, a fluorescence in situ hybridization (FISH) assay was performed to detect the locations of and changes in  Hnf4αos  in mouse hepatocytes after A/R treatment. The results showed that the fluorescence intensity of  Hnf4αos  was markedly enriched in hepatocyte nuclei and significantly elevated in the A/R group compared with the normoxic group ( Fig. 1 F). Therefore, we identified  Hnf4αos  as a novel therapeutic target in the pathogenic process of hepatic I/R injury. Fig. 1 LncRNA  Hnf4αos  is elevated during hepatic I/R injury.  (A)  Heatmaps generated using the RNA expression of members detected by the DEG analysis. The expression of RNAs was visualized in color saturation; the expression level of genes was indicated by the colors (3 mice in the normoxia group and 5 mice in hypoxia group).  (B)  Genomic distance between lncRNAs and correlated with the oxidative stress, inflammatory response and apoptosis genes in KEGG. (The outer ring shows the distribution of the chromosomes of the mouse; The internal lines indicate that the top lncRNA-mRNA pairs)  (C) Hnf4αos  expression was assessed by qRT-PCR in mouse liver I/R models.  (D) Hnf4αos  expression was assessed by qRT-PCR in primary hepatocytes after A/R treatment.  (E)  Levels of cytoplasmic and nuclear  Hnf4αos  in primary hepatocytes.  (F)  The cellular locations and expression changes of  Hnf4αos  were analyzed by RNA-FISH. The scale bar represents 50 μm. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Fig. 1\nLncRNA  Hnf4αos  is elevated during hepatic I/R injury.  (A)  Heatmaps generated using the RNA expression of members detected by the DEG analysis. The expression of RNAs was visualized in color saturation; the expression level of genes was indicated by the colors (3 mice in the normoxia group and 5 mice in hypoxia group).  (B)  Genomic distance between lncRNAs and correlated with the oxidative stress, inflammatory response and apoptosis genes in KEGG. (The outer ring shows the distribution of the chromosomes of the mouse; The internal lines indicate that the top lncRNA-mRNA pairs)  (C) Hnf4αos  expression was assessed by qRT-PCR in mouse liver I/R models.  (D) Hnf4αos  expression was assessed by qRT-PCR in primary hepatocytes after A/R treatment.  (E)  Levels of cytoplasmic and nuclear  Hnf4αos  in primary hepatocytes.  (F)  The cellular locations and expression changes of  Hnf4αos  were analyzed by RNA-FISH. The scale bar represents 50 μm. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)\nTo evaluate the potential effects of  Hnf4αos  on liver damage after hepatic I/R in mice, we altered the expression level of endogenous  Hnf4αos  by tail vein injection with  Hnf4αos  overexpression and downregulation adenoviral vectors ( Supplementary Fig. S2A ). When we knocked down  Hnf4αos  expression in mice, no statistical significance in sham mice was found, and I/R induced tissue necrosis was markedly ameliorated in the liver by silencing  Hnf4αos  expression, whereas,  Hnf4αos  overexpression worsened pathological changes (hemorrhagic change, inflammatory cell infiltration and focal necrosis) in I/R liver tissue ( Fig. 2 A). Additionally, serum aminotransferase (ALT and AST) levels were also significantly decreased in  Hnf4αos  knockdown mice, and ectopic expression of  Hnf4αos  exhibited the opposite effect compared with control mice ( Fig. 2 B and C). Thus, we concluded that  Hnf4αos  exacerbated liver damage induced by HIRI insult. Fig. 2 Hnf4αos  deteriorates liver damage induced by hepatic I/R insult.  (A)  Images (100 × magnification) of H&E-stained liver sections and representative histopathological scores after the transfection of adenovirus vectors. The scale bar represents 200 μm.  (B–C)  Serum levels of aminotransferases (ALT and AST) were detected in the mice subjected to I/R after the transfection of adenovirus vectors.  (D)  Images of H&E-stained liver sections and representative histopathological scores in  Hnf4αos -KO and WT mice. The scale bar represents 200 μm.  (E)  Serum levels of aminotransferases (ALT and AST) were detected in the  Hnf4αos -KO and WT mice subjected to I/R operation. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. Fig. 2\nHnf4αos  deteriorates liver damage induced by hepatic I/R insult.  (A)  Images (100 × magnification) of H&E-stained liver sections and representative histopathological scores after the transfection of adenovirus vectors. The scale bar represents 200 μm.  (B–C)  Serum levels of aminotransferases (ALT and AST) were detected in the mice subjected to I/R after the transfection of adenovirus vectors.  (D)  Images of H&E-stained liver sections and representative histopathological scores in  Hnf4αos -KO and WT mice. The scale bar represents 200 μm.  (E)  Serum levels of aminotransferases (ALT and AST) were detected in the  Hnf4αos -KO and WT mice subjected to I/R operation. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.\nTo obtain more evidence supporting the role of  Hnf4αos  in I/R-induced liver injury, we generated  Hnf4αos -knockout ( Hnf4αos -KO) and  Hnf4αos -wild-type ( Hnf4αos - 10.13039/100010269 WT ) mice ( Supplementary Figs. S2B and C ). Subsequently,  Hnf4αos -KO mice were subjected to a 75-min I/R operation. As expected, histological H&E staining showed considerable amelioration of tissue necrosis levels by  Hnf4αos  knockout ( Fig. 2 D). Moreover,  Hnf4αos -KO mice exhibited reduced release of ALT and AST in serum compared with  Hnf4αos -WT mice ( Fig. 2 E). Of note, serum aminotransferases were significantly lower in the low  Hnf4α-as1  group, suggesting less liver injury and better liver function after partial hepatectomy ( Supplementary Fig. S2D ). Overall, these observations suggest that  Hnf4αos  inhibition plays a protective role in hepatic I/R injury.\nThe sterile inflammatory response plays a pivotal role in I/R injury, and the release of cytokines and chemokines is sustained throughout the entire pathophysiological processes of hepatic I/R. Therefore, we performed RNA-seq with I/R challenged liver samples of WT and  Hnf4αos -KO mice to detect whether  Hnf4αos  can affect liver damage by modulating the inflammatory response. The Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis demonstrated significantly enriched signaling pathways of inflammatory response, in particularly the NF-κB pathway ( Fig. 3 A). Moreover, heatmap of leading-edge enriched pathways showed that  Hnf4αos  ablation mainly affected the expression of NF-κB signaling related molecules ( Fig. 3 B). The ELISA and qRT-PCR analysis suggested sham procedure did not induce basal inflammation changes in mice ( Fig. 3 C and D).  Hnf4αos -KO mice exhibited less inflammatory cytokine/chemokine (TNF-α, IL-1β, IL-6, and MIP-2) release than WT mice in the I/R model ( Fig. 3 C and D). In accordance with the data obtained in vivo, the medium collected from the primary  Hnf4αos -KO hepatocyte culture contained lower levels of cytokines/chemokines ( Supplementary Fig. S3A ). Tissue MPO activity, an indicator of neutrophil infiltration, was dramatically increased following I/R insult in WT mice. In contrast,  Hnf4αos -KO mice exhibited less neutrophil accumulation ( Supplementary Fig. S3B ). Moreover, tissue section immunofluorescence analysis demonstrated fewer Ly6G positive (a neutrophil biomarker) cells when comparing  Hnf4αos -KO versus WT-I/R mice ( Fig. 3 E). Gene set enrichment analysis (GSEA) also indicated that  Hnf4αos  could significantly activate the NF-κB signaling pathway ( Supplementary Fig. S3C ). Subsequently, we found that  Hnf4αos -KO inhibited the translocation of NF-κB from cytoplasm to nuclear during HIRI ( Supplementary Fig. S3D ). Further results showed that NF-κB pathway during the I/R process was obviously reversed in the  Hnf4αos -KO as shown by Western blotting ( Fig. 3 F). As indicated above, we obtained nearly identical results in  Hnf4α-as1  knockdown and overexpression human L02 hepatocytes ( Supplementary Figs. S3E and F ). Fig. 3 Hnf4αos knockout inhibits the inflammatory response during hepatic I/R injury.  (A)  KEGG pathway enrichment analysis of the major biological pathways.  (B)  Heatmap showing expression of inflammatory genes involved in HIRI.  (C)  TNF-α, IL-1β, IL-6 and MIP-2 levels after liver I/R were measured by ELISA.  (D)  Relative mRNA expression of Tnf-α, Il-1β, Il-6 and Mip-2 after liver I/R was examined by qRT-PCR (n = 5).  (E)  Representative immunofluorescence images of the Ly6G after I/R injury and the quantification of Ly6G-positive cell ratio. The scale bar represents 25 μm.  (F)  Western blot analysis of p-IκBα, IκBα, p-p65, and p65 and the relative band density. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. Fig. 3\nHnf4αos knockout inhibits the inflammatory response during hepatic I/R injury.  (A)  KEGG pathway enrichment analysis of the major biological pathways.  (B)  Heatmap showing expression of inflammatory genes involved in HIRI.  (C)  TNF-α, IL-1β, IL-6 and MIP-2 levels after liver I/R were measured by ELISA.  (D)  Relative mRNA expression of Tnf-α, Il-1β, Il-6 and Mip-2 after liver I/R was examined by qRT-PCR (n = 5).  (E)  Representative immunofluorescence images of the Ly6G after I/R injury and the quantification of Ly6G-positive cell ratio. The scale bar represents 25 μm.  (F)  Western blot analysis of p-IκBα, IκBα, p-p65, and p65 and the relative band density. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.\nAn excessive inflammatory response inevitably causes cell death, which is accompanied by varying degrees of liver damage [ 20 ]. Therefore, we further examined the effects of  Hnf4αos  on cell apoptosis. As expected, the I/R model showed a significant elevation in apoptosis, and we found fewer TUNEL-positive cells in liver tissues from  Hnf4αos -KO mice than in liver tissues from  Hnf4αos -WT mice ( Fig. 4 A). Flow cytometry assay showed that  Hnf4αos  depletion reduced the apoptotic levels of hepatocytes subjected to A/R operation compared to  Hnf4αos -WT group ( Fig. 4 B). The results of the caspase-3 activity assay and DNA fragmentation ELISA also suggested dramatic decrease in apoptotic levels with  Hnf4αos  depletion ( Fig. 4 C and D). As shown by qRT-PCR and Western blot, I/R-induced cell death was markedly blunted in the livers of  Hnf4αos  deficient mice, as evidenced by the expression of apoptotic markers (BCL-2, Bax and cleaved caspase-3) ( Fig. 4 E and F). Moreover, less LDH was released from  Hnf4αos  deficient hepatocyte cultures than from control hepatocytes ( Fig. 4 G). The CCK-8 assay results in  Fig. 4 H showed that  Hnf4αos -deficiency enhanced cell viability and promoted cell proliferation in  Hnf4αos -KO mice, compared to control mice. In line with our observations in primary mouse hepatocytes,  Hnf4α-as1 -knockdown in human L02 hepatocytes also alleviated cell apoptosis and  Hnf4α-as1 -overexpression had the opposite effects ( Supplementary Fig. S4A ). Fig. 4 Hnf4αos  depletion protects hepatocytes from hepatic I/R injury in vivo and in vitro.  (A)  Representative images of liver sections stained by TUNEL and the quantification of the TUNEL-positive cell ratio. The scale bar represents 25 μm.  (B)  Cell apoptosis determined by flow cytometry and the quantification of the apoptotic cells.  (C – D)  Caspase-3 activity and DNA fragmentation in mouse liver extracts were determined by ELISA.  (E)  Relative mRNA expression of Bcl2 and Bax.  (F)  Western blot analysis of BCL2, BAX, c-CASP and relative band density.  (G)  LDH release from hepatocytes was measured after A/R treatment.  (H)  Cell viability was determined at different timepoints after A/R treatment by CCK-8 assay. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. Fig. 4\nHnf4αos  depletion protects hepatocytes from hepatic I/R injury in vivo and in vitro.  (A)  Representative images of liver sections stained by TUNEL and the quantification of the TUNEL-positive cell ratio. The scale bar represents 25 μm.  (B)  Cell apoptosis determined by flow cytometry and the quantification of the apoptotic cells.  (C – D)  Caspase-3 activity and DNA fragmentation in mouse liver extracts were determined by ELISA.  (E)  Relative mRNA expression of Bcl2 and Bax.  (F)  Western blot analysis of BCL2, BAX, c-CASP and relative band density.  (G)  LDH release from hepatocytes was measured after A/R treatment.  (H)  Cell viability was determined at different timepoints after A/R treatment by CCK-8 assay. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.\nBased on the GEO data-set ( GSE15891 ), we found the differentially expressed genes (DEGs) ( Fig. 5 A) are closely related to the regulation of cell death, oxidative and anti-inflammatory response according to the Gene Ontology (GO) analysis ( Fig. 5 B). Moreover, we established a module by bioinformatic methods to evaluate the potential correlation between the DEGs and differential expressed lncRNAs. The lncRNA-mRNA interaction network ( Fig. 5 C) surprisingly revealed a close correlation between  Hnf4αos  and PGC1α. We previously reported that PGC1α protected the liver from I/R injury by attenuating hepatocyte death, reducing cytokine/chemokine release and alleviating oxidative stress [ 13 ]. GSEA also demonstrated that most genes affected by PGC1α overexpression were involved in the KEGG apoptosis pathway. More importantly, a dramatically negative correlation was found between  Hnf4αos  and PGC1α pathway related molecules ( Fig. 5 D). Specifically, in  Fig. 5 E, the module enriched in multiple cell death, oxidative stress and inflammatory pathways also showed a high degree of correlation with  Hnf4αos  and PGC1α expression. Thus, we confirmed an obviously negative association between  Hnf4αos  and PGC1α by Western blot ( Fig. 6 A). Our previous study found that PGC1α can protect the liver against I/R insult by accelerating the clearance of ROS. Therefore, we hypothesized that  Hnf4αos -KO ameliorates liver damage in the I/R process by scavenging accumulated ROS. Subsequently, we detected ROS levels by dihydroethidium staining (DHE) and DHE staining showed that in the livers of  Hnf4αos -KO mice, intracellular concentrations of ROS were markedly decreased compared with those in control mice subjected to I/R operation ( Fig. 6 B). As indicators of oxidative stress damage, MDA and 4-HNE contents were tested in I/R-treated liver tissues. In line with the results of DHE staining,  Hnf4αos  knockout abrogated the I/R-induced increase in MDA/4-HNE contents and resulted in lower MDA/4-HNE contents ( Fig. 6 C;  Supplementary Fig. S5 ). Next, we speculated whether the activities of ROS scavenging enzymes were increased, which were induced by  Hnf4αos  knockout-mediated PGC1α upregulation. The hepatic activities of ROS scavenging enzymes (SOD, CAT and GPX) were increased in the KO groups compared with the WT mice following the I/R operation ( Fig. 6 D). In line with the activities of antioxidative enzymes, the mRNA levels of  Sod1 ,  Sod2 ,  Cat  and  Gpx1  were dramatically decreased after mice were subjected to the I/R procedure. However,  Hnf4αos -KO enhanced the expression of those enzymes in the I/R model compared to that in WT mice ( Fig. 6 E). We then constructed an shPGC1α adenovirus and transferred PGC1α-deficient vectors into  Hnf4αos -KO mice and primary hepatocytes ( Supplementary Figs. S6A and B ). Reversibility experiments ensured that PGC1α knockdown abrogated the reduced oxidative stress damage induced by  Hnf4αos -KO and that  Hnf4αos -KO-mediated protection against hepatic I/R injury was also reversed by PGC1α deficiency ( Fig. 6 F–L;  Supplementary Figs. S6C–E ). Fig. 5 PGC1α is the potential target of  Hnf4αos .  (A)  Heatmaps generated using the RNA expression of members detected by the DEG (3 mice in the nornoxia group and 5 mice in hypoxia group).  (B)  Gene Ontology (GO) analysis for DEGs that correlated with cell death, oxidative stress and inflammatory response.  (C)  lncRNA-mRNA interaction module of the network indicates that PGC1α and  Hnf4αos  were potentially correlated. Blue nodes represent mRNAs, red nodes represent lncRNAs, and lines indicate interactions.  (D)  GSEA of apoptosis gene signatures in PGC1α enrichment groups and coexpressed genes of PGC1α and  Hnf4αos  function determined by GSEA respectively.  (E)  Gene regulated genes by PGC1α and  Hnf4αos  related to the cell death, oxidative stress and inflammatory response pathways. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Fig. 5 Fig. 6 PGC1α mediates  Hnf4αos  function in hepatic I/R injury.  (A)  Western blot analysis of PGC1α in  Hnf4αos -KO mice and  Hnf4αos -KO hepatocytes after I/R and A/R treatment and relative band density.  (B)  Representative images of DHE-stained liver cryosections from  Hnf4αos -KO mice after I/R injury. The scale bar represents 50 μm  (C)  The MDA content after liver I/R injury.  (D)  The activities of SOD, CAT and GPX in the  Hnf4αos -KO mice after I/R injury.  (E)  The relative expression levels of  Sod1 ,  Sod2 ,  Cat  and  Gpx1  mRNA.  (F)  Representative images of DHE-stained liver cryosections from  Hnf4αos -KO mice subjected to Ad-shPGC1α after I/R injury. The scale bar represents 50 μm  (G – H)  MDA and 4-HNE contents from  Hnf4αos -KO mice subjected to Ad-shPGC1α after liver I/R injury.  (I)  The relative expression levels of  Sod1 ,  Sod2 ,  Cat  and  Gpx1  mRNA from  Hnf4αos -KO mice subjected to Ad-shPGC1α.  (J)  Representative images of H&E-stained liver sections. The scale bar represents 200 μm.  (K)  Representative images of liver sections stained by TUNEL. The scale bar represents 25 μm.  (L)  Western blot analysis of PGC1α, NF-κB and apoptosis related genes and relative band density. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. Fig. 6\nPGC1α is the potential target of  Hnf4αos .  (A)  Heatmaps generated using the RNA expression of members detected by the DEG (3 mice in the nornoxia group and 5 mice in hypoxia group).  (B)  Gene Ontology (GO) analysis for DEGs that correlated with cell death, oxidative stress and inflammatory response.  (C)  lncRNA-mRNA interaction module of the network indicates that PGC1α and  Hnf4αos  were potentially correlated. Blue nodes represent mRNAs, red nodes represent lncRNAs, and lines indicate interactions.  (D)  GSEA of apoptosis gene signatures in PGC1α enrichment groups and coexpressed genes of PGC1α and  Hnf4αos  function determined by GSEA respectively.  (E)  Gene regulated genes by PGC1α and  Hnf4αos  related to the cell death, oxidative stress and inflammatory response pathways. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)\nPGC1α mediates  Hnf4αos  function in hepatic I/R injury.  (A)  Western blot analysis of PGC1α in  Hnf4αos -KO mice and  Hnf4αos -KO hepatocytes after I/R and A/R treatment and relative band density.  (B)  Representative images of DHE-stained liver cryosections from  Hnf4αos -KO mice after I/R injury. The scale bar represents 50 μm  (C)  The MDA content after liver I/R injury.  (D)  The activities of SOD, CAT and GPX in the  Hnf4αos -KO mice after I/R injury.  (E)  The relative expression levels of  Sod1 ,  Sod2 ,  Cat  and  Gpx1  mRNA.  (F)  Representative images of DHE-stained liver cryosections from  Hnf4αos -KO mice subjected to Ad-shPGC1α after I/R injury. The scale bar represents 50 μm  (G – H)  MDA and 4-HNE contents from  Hnf4αos -KO mice subjected to Ad-shPGC1α after liver I/R injury.  (I)  The relative expression levels of  Sod1 ,  Sod2 ,  Cat  and  Gpx1  mRNA from  Hnf4αos -KO mice subjected to Ad-shPGC1α.  (J)  Representative images of H&E-stained liver sections. The scale bar represents 200 μm.  (K)  Representative images of liver sections stained by TUNEL. The scale bar represents 25 μm.  (L)  Western blot analysis of PGC1α, NF-κB and apoptosis related genes and relative band density. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.\nTo determine how  Hnf4αos  manipulates hepatocyte viability by regulating PGC1α, we further conducted an in-depth study of the structural features of  Hnf4αos .  Hnf4αos  is a natural antisense transcript (NAT) of  Hnf4α  known for its transcriptional regulation of several hepatic genes. As reported previously, antisense lncRNAs are used to bind to the respective sense strand mRNA to form a duplex strand, which enhances the stability of the latter mRNA [ [21] ,  [22] ,  [23] ]. We further explored the mRNA and protein levels of Hnf4α accompanied by  Hnf4αos  alteration. As shown in  Fig. 7 A and B, downregulated  Hnf4αos  expression significantly decreased the mRNA and protein levels of Hnf4α. Conversely,  Hnf4αos  overexpression enhanced the expression levels of Hnf4α. Then, we constructed Hnf4α overexpression and Hnf4α knockdown adenovirus vectors ( Supplementary Figs. S7A–B ). However, the variations in Hnf4α expression had no effects on the  Hnf4αos  transcript ( Fig. 7 C). To determine whether  Hnf4αos  regulated the stability of  Hnf4α  mRNA, we performed an RNA stability assay.  Hnf4αos -KO and  Hnf4αos- overexpressing hepatocytes were treated with actinomycin D (ActD) to inhibit mRNA transcription. qRT-PCR analysis showed that  Hnf4αos  downregulation markedly shortened the half-life of  Hnf4α  mRNA and that  Hnf4αos  overexpression elevated the level of  Hnf4α  mRNA ( Fig. 7 D). These findings indicate that  Hnf4αos  positively regulates  Hnf4α  mRNA expression. Fig. 7 Hnf4αos  promotes the stability of  Hnf4α  mRNA.  (A)  The relative expression levels of Hnf4α mRNA in Ad- Hnf4αos  and Ad-sh Hnf4αos  cells.  (B)  Western blot analysis of Hnf4α in Ad- Hnf4αos  and Ad-sh Hnf4αos  cells and relative band density.  (C)  The relative expression levels of  Hnf4αos  RNA in Ad-Hnf4α and Ad-shHnf4α cells.  (D)  After treatment with ActD (5 g/ml), the stability of  Hnf4α  and  Gapdh  mRNA in the cells transfected with Ad- Hnf4αos , Ad-sh Hnf4αos  and the respective control vectors was determined by qRT-PCR at different timepoints.  (E)  Schematic representation of the  Hnf4αos / Hnf4α  locus.  (F) Hnf4α  mRNA levels measured by qRT-PCR followed by ribonuclease protection assay.  (G)  The interaction between  Hnf4α  and biotin- Hnf4αos  was detected by a biotin RNA pulldown assay followed by qRT-PCR.  (H)  The interaction between  Hnf4α  and  Hnf4αos  was detected by TRAP assay.  (I)  Representative images of H&E-stained liver sections from  Hnf4αos -KO mice subjected to Ad-Hnf4α after liver I/R injury. The scale bar represents 200 μm.  (J)  Representative images of liver sections stained by TUNEL. The scale bar represents 25 μm.  (K)  Western blot analysis of Hnf4α, NF-κB and apoptosis related genes and relative band density. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. Fig. 7\nHnf4αos  promotes the stability of  Hnf4α  mRNA.  (A)  The relative expression levels of Hnf4α mRNA in Ad- Hnf4αos  and Ad-sh Hnf4αos  cells.  (B)  Western blot analysis of Hnf4α in Ad- Hnf4αos  and Ad-sh Hnf4αos  cells and relative band density.  (C)  The relative expression levels of  Hnf4αos  RNA in Ad-Hnf4α and Ad-shHnf4α cells.  (D)  After treatment with ActD (5 g/ml), the stability of  Hnf4α  and  Gapdh  mRNA in the cells transfected with Ad- Hnf4αos , Ad-sh Hnf4αos  and the respective control vectors was determined by qRT-PCR at different timepoints.  (E)  Schematic representation of the  Hnf4αos / Hnf4α  locus.  (F) Hnf4α  mRNA levels measured by qRT-PCR followed by ribonuclease protection assay.  (G)  The interaction between  Hnf4α  and biotin- Hnf4αos  was detected by a biotin RNA pulldown assay followed by qRT-PCR.  (H)  The interaction between  Hnf4α  and  Hnf4αos  was detected by TRAP assay.  (I)  Representative images of H&E-stained liver sections from  Hnf4αos -KO mice subjected to Ad-Hnf4α after liver I/R injury. The scale bar represents 200 μm.  (J)  Representative images of liver sections stained by TUNEL. The scale bar represents 25 μm.  (K)  Western blot analysis of Hnf4α, NF-κB and apoptosis related genes and relative band density. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.\nIn the case of the  Hnf4αos/Hnf4α  pair, complementarity was noted in both transcripts ( Fig. 7 E). To determine the existence of a sense-antisense RNA duplex, a ribonuclease protection assay (RPA) was performed and showed that the complementary region was protected from degradation by RNase, indicating an RNA duplex between lncRNA  Hnf4αos  and  Hnf4α  mRNA ( Fig. 7 F). Furthermore, the biotin-labeled RNA pulldown assay and tagged RNA affinity purification (TRAP) assay revealed a strong interaction between  Hnf4αos  and endogenous  Hnf4α  mRNA ( Fig. 7 G and H). We noticed that enhanced expression of Hnf4α worsened liver injury ( Fig. 7 I–K,  Supplementary Figs. S7C and E ) and activated a sterile inflammatory response ( Fig. 7 K,  Supplementary Fig. S7D ), as evidenced by more severe tissue necrosis and cytokine/chemokine release, which could be ameliorated by  Hnf4αos -KO. Collectively, these data support the conclusion that  Hnf4αos  increased the stability of  Hnf4α  mRNA, which was modulated by the duplex of  Hnf4αos/Hnf4α .\nTo further confirm the exact mechanism through which  Hnf4αos  regulated PGC1α expression, we speculated that Hnf4α exerted a directive transcriptional inhibitory effect on PGC1α by acting as a transcription factor (TF). In support of our hypothesis, we analyzed the PGC1α promoter sequences using the UCSC, JASPAR, SWISSREGULON and PROMO algorithms and surprisingly found that the promoter region of PGC1α has a candidate binding site for TF-Hnf4α ( Fig. 8 A). The luciferase reporter assay demonstrated no relationship between Hnf4α and the transcriptional activity of PGC1α ( Fig. 8 B). Numerous reports have shown that miR-23a is a key regulator of PGC1α expression [ [24] ,  [25] ,  [26] ], and we found a physical interaction between miR-23a and PGC1α through the miRDB, RNAinter, TargetScan and miRmap databases ( Fig. 8 C). The luciferase reporter assay confirmed that miR-23a was a negative regulator of PGC1α ( Fig. 8 D). Then, we performed qRT-PCR to detect the RNA level of miR-23a between  Hnf4αos  and  Hnf4α  ( Supplementary Fig. S8 ). To confirm that miR-23a contributes to the function of PGC1α in hepatic I/R injury, we constructed miR-23a mimics and inhibitors. Western blot analysis showed that miR-23a and Hnf4α deficiency dramatically upregulated the protein levels of PGC1α, conversely, miR-23a/Hnf4α overexpression suppressed PGC1α protein expression ( Fig. 8 E). Given that the considerable lncRNA  Hnf4αos  enhances the stability of Hnf4α, we speculated whether TF-Hnf4α mediated the transcription of miR-23a and subsequently attenuated the expression of PGC1α. Intriguingly, based on the prediction by the database, we found that Hnf4α binding sites in the promoter of miR-23a and revealed that the transcription of miR-23a was dramatically activated by TF-Hnf4α ( Fig. 8 F and G). Consistently, nuclear extracts were obtained and used for an electrophoretic mobility shift assay (EMSA), and the results identified marked DNA-protein binding activity in mouse primary hepatocytes ( Fig. 8 H). Furthermore, chromatin immunoprecipitation (ChIP) assays provided evidence for the direct interaction of Hnf4α with the miR-23a promoter ( Fig. 8 I). Together, the data above showed that a significant interaction between the promoter region of miR-23a and TF-Hnf4α. Fig. 8 Hnf4α enhances miR-23a transcription by binding to its promoter region.  (A)  Hnf4α consensus and its potential binding sites on the PGC1α promoter.  (B)  The results of the luciferase reporter assay.  (C)  Complementary WT and Mut sequence alignment of miR-23a and PGC1α.  (D)  The results of the luciferase reporter assay.  (E)  Western blot analysis of PGC1α and Hnf4α with the transfection of Ad-Hnf4α, Ad-shHnf4α, miR-23a inhibitor and miR-23a mimics and relative band density.  (F)  Hnf4α consensus and its potential binding sites on the miR-23a promoter.  (G)  The results of the luciferase reporter assay.  (H)  EMSA was performed with nuclear extracts and radiolabeled probes encompassing the candidate Hnf4α-binding sequence on the miR-23a promoter.  (I)  ChIP assay showing the binding of Hnf4α to the miR-23a promoter. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. Fig. 8\nHnf4α enhances miR-23a transcription by binding to its promoter region.  (A)  Hnf4α consensus and its potential binding sites on the PGC1α promoter.  (B)  The results of the luciferase reporter assay.  (C)  Complementary WT and Mut sequence alignment of miR-23a and PGC1α.  (D)  The results of the luciferase reporter assay.  (E)  Western blot analysis of PGC1α and Hnf4α with the transfection of Ad-Hnf4α, Ad-shHnf4α, miR-23a inhibitor and miR-23a mimics and relative band density.  (F)  Hnf4α consensus and its potential binding sites on the miR-23a promoter.  (G)  The results of the luciferase reporter assay.  (H)  EMSA was performed with nuclear extracts and radiolabeled probes encompassing the candidate Hnf4α-binding sequence on the miR-23a promoter.  (I)  ChIP assay showing the binding of Hnf4α to the miR-23a promoter. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.\nAs shown by confocal microscopy examination, both  Hnf4αos  and miR-23a levels were increased in A/R-treated cells compared with normoxic cells by dual-RNA FISH detection ( Fig. 9 A). Moreover, the primary hepatocytes subjected to  Hnf4αos -KO exhibited almost no red/green fluorescence signals, while the fluorescence signals of WT-cells were much stronger ( Fig. 9 B). Further experiments validated that miR-23a overexpression attenuated the protective effects of  Hnf4αos -KO on liver I/R injury ( Fig. 9 C–E and  Supplementary Figs. S9A–C ). Importantly, both Hnf4α and miR-23a also abrogated the antioxidative effects induced by  Hnf4αos -KO, and miR-23a knockdown suppressed the oxidative activation of Hnf4α overexpression as demonstrated by  Fig. 9 F–I. Finally, miR-23a deficiency ameliorated liver damage and the inflammatory response induced by Hnf4α overexpression ( Supplementary Figs. S9D–G ). These data suggest that Hnf4α mediates the suppressive effect of miR-23α on PGC1α. Fig. 9 miR-23a exacerbates liver damage and oxidative stress induced I/R injury.  (A)  The cellular expression of  Hnf4αos  and miR-23a was analyzed by dual RNA-FISH after A/R treatment. The scale bar represents 10 μm.  (B)  The cellular expression of  Hnf4αos  and miR-23a was analyzed by Dual RNA-FISH in  Hnf4αos -KO hepatocytes. The scale bar was 20 μm.  (C)  Representative images of H&E-stained liver sections from  Hnf4αos -KO mice subjected to miR-23a mimics after liver I/R injury and the quantification of histological score. The scale bar represents 200 μm.  (D)  Representative images of liver sections stained by TUNEL after I/R injury and the quantification of the TUNEL-positive cell ratio. The scale bar was 25 μm.  (E)  Western blot analysis of NF-κB and apoptosis related genes and relative band density.  (F–I)  miR-23a inhibitor reverses the levels of oxidative stress induced by Hnf4α overexpression.  (J)  Mechanism involved in protective effects of  Hnf4αos -KO after liver subjected to I/R insult. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. Fig. 9\nmiR-23a exacerbates liver damage and oxidative stress induced I/R injury.  (A)  The cellular expression of  Hnf4αos  and miR-23a was analyzed by dual RNA-FISH after A/R treatment. The scale bar represents 10 μm.  (B)  The cellular expression of  Hnf4αos  and miR-23a was analyzed by Dual RNA-FISH in  Hnf4αos -KO hepatocytes. The scale bar was 20 μm.  (C)  Representative images of H&E-stained liver sections from  Hnf4αos -KO mice subjected to miR-23a mimics after liver I/R injury and the quantification of histological score. The scale bar represents 200 μm.  (D)  Representative images of liver sections stained by TUNEL after I/R injury and the quantification of the TUNEL-positive cell ratio. The scale bar was 25 μm.  (E)  Western blot analysis of NF-κB and apoptosis related genes and relative band density.  (F–I)  miR-23a inhibitor reverses the levels of oxidative stress induced by Hnf4α overexpression.  (J)  Mechanism involved in protective effects of  Hnf4αos -KO after liver subjected to I/R insult. n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.\n\nHIRI is the most important effector in liver surgery, particularly in liver transplantation. In the present study, we found a differentially expressed lncRNA –  Hnf4αos  during HIRI progression in both human and mouse models. Knocking out  Hnf4αos  in hepatocytes significantly suppressed the oxidative stress – induced hepatic injury and inhibited the inflammatory response during HIRI both in vitro and in vivo. Using the integrated approaches of bioinformatic analysis, we identify the potential interaction of  Hnf4αos  and PGC1α, and  Hnf4αos  facilitated the RNA decay of PGC1α by ceRNA function. Thus,  Hnf4αos  could be a promising therapeutic target of HIRI.\nOxidative stress – induced liver injury plays dominant roles during HIRI progression. The production of ROS caused by the oxidative stress response triggers peroxidation reactions, which activate the apoptotic pathway and decrease hepatocyte viability in hepatic I/R injury [ 27 , 28 ]. Therefore, regulation of ROS metabolism is expected to have the potential to effectively protect the liver against I/R injury. In the current study, we found that the reduced  Hnf4αos  level exhibited a significant antioxidative effects by regulating the balance of ROS scavenging and accumulation systems. Considering that PGC1α is key mediator of ROS metabolism [ 13 , 18 ], we speculated that  Hnf4αos  regulated the oxidative stress in HIRI by targeting PGC1α. To verify our hypothesis,  Hnf4αos -KO mice were generated.  Hnf4αos  deficiency in vivo and in vitro reduces the degree of hepatic I/R and improves hepatic function in mice by PGC1α elevation-mediated ROS scavenging compared to WT conditions.\nFunctionally, lncRNAs can bind not only to proteins but also to DNA and RNA, rendering lncRNAs a crucial factor in protein-nucleic acid/nucleic acid-nucleic acid networks. Several studies, including ours, have provided strong evidence that NATs regulate the expression of their sense protein-coding mRNAs [ 21 , 22 ]. Here, our findings demonstrated that  Hnf4αos  and  Hnf4α  formed an RNA – RNA duplex and further promoted  Hnf4α  mRNA stability, which consequently enhanced the protein level of Hnf4α as shown by Western blot.\nHnf4α generally functions as a transcription factor in the liver and has been reported to play prominent roles in cell proliferation, cell differentiation, lipid metabolism and gluconeogenesis [ [29] ,  [30] ,  [31] ,  [32] ]. Several studies have revealed that Hnf4α is a key regulator in inhibiting hepatocyte proliferation. Walesky et al. found that hepatocyte-specific depletion of Hnf4α induced increased levels of cell proliferation. Further microarray analysis demonstrated that a significant number of genes known to be promitogenic were upregulated by Hnf4α-deficiency [ 33 ]. It has been reported that Hnf4α promoted the transcriptional activity of ASK1, which is a typical proapoptotic mediator in MAPK pathway [ 34 ] and Mai et al. confirmed the antiapoptotic potential of Hnf4α-deficiency in endometriosis [ 35 ]. Furthermore, a study related to viral hepatitis suggested that knocking down Hnf4α markedly inhibited HBV RNA transcripts and respective DNA replication intermediates, which played a key role in delaying the progression of HBV-induced hepatitis [ 36 ]. Although Hnf4α exerts antineoplastic activity in HCC, Hnf4α was reported to act as an oncogene in gastrointestinal adenocarcinomas and pancreatic cancer [ 37 , 38 ], indicating multiple roles of Hnf4α. In our study, we found that Hnf4α served as a TF binding site in the promoter of miR-23a and subsequently further activated its transcription, showing antiproliferative and proapoptotic effects in HIRI accompanied by  Hnf4αos  depletion.\nAlthough our clinical data showed a downward trend of  Hnf4α-as1  in patients who underwent partial liver resections,  Hnf4α  and miR-23a were also downregulated simultaneously during I/R insult, which may be due to the species differences between humans and mice [ 39 ] ( Supplementary Fig. S10 ). More importantly,  Hnf4α-as1  deficiency also exerted hepatoprotective effects on the process of HIRI, and  Hnf4α-as1  overexpression had the opposite effects. Consequently, from this perspective, clinical therapeutic strategies targeting  Hnf4α-as1  can be reasonably established.\n\nIn conclusion, our findings demonstrate a strategy to manipulate PGC1α activity by  Hnf4αos . Specifically,  Hnf4αos -mediated stabilization of  Hnf4α  mRNA reverses the protective effect of PGC1α by upregulating miR-23a expression, leading to a reduction in the scavenging levels of ROS and exacerbation of hepatic I/R injury. Thus, targeting  Hnf4α-as1  may provide potential clinical benefits for liver I/R injury.\n\nNo potential conflicts of interest were disclosed.\n\nThis work was jointly supported by grants from the Outstanding Youth Training Fund from Academician Yu Weihan of  10.13039/100010722 Harbin Medical University  (2014),  Harbin Medical University Postgraduate Innovation and Practical Research Project  (YJSCX2020-28HYD), Scientific Foundation of the First Affiliated Hospital of  10.13039/100010722 Harbin Medical University  (2019L01, HYD2020JQ0007, HYD2020JQ0011),  Heilongjiang Postdoctoral Foundation  (LBH-Z11066, LBH-Z12201 and LBH-Q17097),  10.13039/501100002858 China Postdoctoral Science Foundation  (2012M510990, 2012M520769 and 2013T60387),  10.13039/501100005046 Natural Science Foundation of Heilongjiang Province  of China (LC2018037) and the  10.13039/100014717 National Natural Scientific Foundation of China  (81100305, 81470876 and 81270527).","source_license":"CC-BY-4.0","license_restricted":false}