The healthy male C57BL/6J mice (6 weeks) were purchased from Guangdong Experimental Animal Center (Guangzhou, China). Keep food and water free in an environment with regular light time (12 h/day) and relative humidity of 50%–70%. The mice acclimated to the environment for 7 days. Then, seventeen mice were randomized into two groups, nine mice were injected intraperitoneally with thioacetamide (HE group) and eight mice from the control group (Ctrl group). Dilute the thioacetamide powder (TAA; Sigma Aldrich, United States) to 10 mg/ml with normal saline when it will be used. The drug dosage was determined according to the body weight of the mice at the time of administration. The HE group received 100 mg/kg thioacetamide intraperitoneal injections daily on the first and second days, and then 50 mg/kg TAA daily from the third to the seventh day (Sun et al., 2019). Control group were treated with an equal volume of normal saline. Following the completion of the model building, the successful construction of HE mice were confirmed by liver pathological examination and behavioral examination. Hippocampus tissues from HE mice and ctrl mice (n = 3) were used for the analysis of RNAs, miRNAs and mRNAs. The animal experiments were approved by the ethics committee of Jinan University.

For mRNA and lncRNA detection, reverse transcription was performed with the Reverse Transcription Kit (Invitrogen, United States) according to the manufacturer’s protocol. Real-time quantitative PCR analysis was performed using the talent real-time PCR kit (TIAGEN, China). GAPDH was used as an endogenous control for lncRNAs and mRNA. PCR cycling conditions were conducted as follows: 95°C for 3 min; 40 cycles of 95°C for 5 s, and 60°C for 15 s. To analyze miRNA expression, cDNA synthesis was performed via the miRNA First-Strand cDNA Synthesis kit (TIANGEN, China). The reaction conditions were as follows: 42°C for 60 min, 95°C for 5 min. miRNA stem-loop primers were designed by BGI (Shenzhen, China), and the Real-Time primers were using Primer Premier 5.0 software based on the mature miRNA sequences. U6 small nuclear RNA served as endogenous controls. The quantification of miRNA was performed using the TaqMan™ miRNA Assay (cat. no. 4427975; Thermo Fisher Scientific, Inc.) RT-qPCR and thermocycling conditions were conducted as follows: 95°C 2 min; 40 cycles 95°C 15 s, 60°C 15 s and 68°C 30 s. The PCR amplification was performed in a GeneAmp PCR system 9700 Thermocycler (Applied Biosystems). SIX3OS1-shRNA and NC-shRNA lentiviruses were purchased from Guangzhou Aiji Biotechnology Co., Ltd. The primer sequences are detailed in Supplementary Table S1.

To assess functional enrichment, Gene Ontology (GO) Biological Processes term and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses of mRNAs in the ceRNA network [The ceRNA hypothesis suggests that some RNAs, as ceRNAs, can regulate downstream mRNA’s expression by combining shared miRNAs. This hypothesis describes that ceRNAs can switch the function of target miRNAs by competing for the mRNA co-binding sites on the target miRNAs (Karreth and Pandolfi, 2013)] were performed using Cytoscape. GO enrichment analysis provides all GO terms that significantly enriched in ceRNAs comparing to the genome background, and filter the ceRNAs that correspond to biological functions. Firstly, all ceRNAs were mapped to GO terms in the Gene Ontology database (http://www.geneontology.org/), gene numbers were calculated for every term, significantly enriched GO terms in ceRNAs comparing to the genome background were defined by hypergeometric test. Genes usually interact with each other to play roles in certain biological functions. Pathway-based analysis helps to further understand genes biological functions. KEGG is the major public pathway-related database (http://www.kegg.jp/kegg/). Pathway enrichment analysis identified significantly enriched metabolic pathways or signal transduction pathways in ceRNAs compared with the whole genome background.

The potential functions of the expressed genes could be inferred through interaction networks (Salmena et al., 2011). To determine the correlation between miRNA-mRNA or miRNA-lncRNA. Firstly, the targeting relationship between miRNAs and candidate ceRNAs (lncRNA, mRNA) and the negative correlation relationship between expression levels were analyzed; then the positive correlation relationship between the expression levels of candidate ceRNAs was analyzed. Finally, candidate ceRNAs and their shared miRNA pairs for constructing ceRNA regulatory networks are obtained, that is, the lncRNA-miRNA-mRNA relationship pair. The networks were visualized using Cytoscape software (v3.6.0) (http://www.cytoscape.org/).

The culture of neurons was as previously described (Wang et al., 2018). In short, the hippocampus was isolated from P0 mice, stored in ice-cold Hank’s balanced salt solution, and incubated with 20 units/ml papain at 37°C for 30 min. The dissociated cells were suspended in the plating medium [DMEM/F12 (gibco) + 10% FBS (gibco)] and spread on the coverslips with PDL in a 24-well plate at a density of 30,000–60,000/cm². After the initial incubation for 4 h, changed the medium to a maintenance medium [Neurobasal-Aedium (1x) (gibco) + B27 (gibco) + 1% GlutaMax (gibco) + 1% penicillin-streptomycin (Meilunbio)]. Neurons were kept in incubators with 37°C, 5% CO2. Replace half of the medium every 3 days.

Neurons were washed with pre-cooled phosphate-buffered saline (PBS) once, following fixed with 4% paraformaldehyde (Meilunbio) for 1 h at room temperature. 0.1% Triton X-100 permeabilized neurons for three times (7 min each); then incubated with 3% Bovine Serum Albumin (Genview; FA016) for 1 h at room temperature; washed with PBS for three times (5 min each); and then incubated with anti-GFP (1:10,000; Abcam Ab290) in a humidified chamber overnight at 4°C; Neurons were incubated with Alexa FluorTM 488 secondary antibody (1:500, Thermo Fisher Scientific) was incubated for 1 h at room temperature the next day; after immersion for 3 times, it was mounted with DAPI-containing mounting medium. Observe under a laser confocal microscope, and use ImageJ software to determine the parameters.

The resistance of the glass electrode is 3～5 MΩ. When the electrode tip forms a GΩ seal with the neuron, the negative pressure sucks and breaks the cell, and then performs voltage clamp recording in the whole cell mode. The collected current was amplified by the amplifier multiclamp 700B, the software was Clampex10.5, and the data obtained were sorted and analyzed by Clampfit and miniAnalysis. The extracellular fluid contains NaCl 128 mmol/L, KCl 5 mmol/L, glucose 30 mmol/L, CaCl2 2 mmol/L, Hepes 25 mmol/L, MgCl2 1 mmol/L, and the pH is 7.3. Record microinhibitory synapses for post-current (mIPSC), add tetrodotoxin (TTX, 1 μM), 2-amino-5-phosphonovaleric acid (APV, 25 μM), and 6-cyano-7-nitroquine to the above extracellular fluid. Oxaline-2,3-dione (CNQX, 20 μM), intracellular fluid containing CsCl 140 mmol/L, EGTA 10 mmol/L, Hepes 5 mmol/L, CaCl2 2 mmol/L, MgATP 2 mmol/L, NaGTP 0.3 mmol/L, QX-314 5 mmol/L. When recording the tiny excitatory postsynaptic current (mEPSC), add TTX (1 μM) and PTX (100 μM) to the extracellular fluid; the intracellular fluid contains K-gluconate 125 mmol/L, ethylene glycol ditetraacetic acid (EGTA) 5 mmol/L, KCl 10 mmol/L, Tris-phosphocreatine 10 mmol/L, Hepes 10 mmol/L, NaGTP 0.5 mmol/L, MgATP 4 mmol/L, pH 7.3.

GraphPad Prism 8.0.1 (Graphpad Software Inc.) was used for data analysis and graphing. Data were presented as mean ± s.e.m. A two-tailed Student’s t-test was used to evaluate the differences between groups. p < 0.05 was considered statistically significant.

Before RNA-seq, HE mouse models were constructed and identified (Figure 1A). During the modeling period, the weight of HE mice increased slowly compared with control mice (Figure 1B). The lack of significant increase in body weight in HE mice indicated that TAA induced impaired liver function in mice, leading to decreased appetite, reduced diet, and malnutrition in HE mice, resulting in no significant increase in body weight. Compared to the control group, liver tissue in the HE group showed characteristic focal necrosis and inflammatory infiltration (Figure 1C). The serum ammonia levels of the two groups of mice were also detected. The results showed that the serum ammonia level of the HE mice increased significantly (Figure 1D). To evaluate the behavioral changes of HE mice, open field test was first performed (Figure 1E). The results showed that compared with control mice, there was no statistical difference in the total distance of movement, the distance moved in the central area, the number of passes through the center area, and the time spent in the center of HE group (Figures 1F–I). It suggested that HE mice had no motor dysfunction and no anxiety-like alteration. In addition, the Morris water maze test was used to examine the spatial learning and memory of mice (Figure 1J). Our results showed that compared with control group, HE mice took longer time to find the platform from the fourth day of training (p < 0.01; Figure 1K), and after the platform removal, platform crossing times and time spent at the target quadrant of HE group was significantly less than those of the control group (p < 0.05; Figures 1L,M). In the pre-experiment, we also conducted the elevated plus maze test, the Y-maze test, the novel-object recognition and the three-Chambered Social Test. The results showed that there were no differences in short-term memory, social and motor functions, and no anxiety-like alteration between the two groups of mice (Supplementary Figure S1). These results indicated that HE mice’s spatial learning and memory abilities were impaired, which corresponded to the clinical manifestations of hepatic encephalopathy patients. The evidence above data suggests that we have successfully established mouse models of hepatic encephalopathy. Subsequently, the hippocampal tissues of HE mice and control mice were used for RNA-seq.

We performed a whole transcriptomic analysis of the HE and the control groups to evaluate RNA expression differences (The data used herein has been deposited in NCBI’s Gene Expression Omnibus and are accessible through https://www.ncbi.nlm.nih.gov/bioproject/PRJNA804405). RNA sequencing of six cDNA libraries yielded more than 60 million original reads, most of which were clean reads, and more than 99.76% of clean reads were completely mapped to the reference mouse genome (Table 1). RNA-seq identified 30,450 lncRNAs. The average length of lncRNAs was 1121 bp, of which 86% were shorter than 2000 bp (Figure 2A). The classification of lncRNA included 11,028 (36%) intergenic, 13,951 (46%) sense, 979 (3%) bidirectional, 3,443 (11%) antisense, and 478 (2%) intronic and 571 (2%) other IncRNAs (Figure 2B). All detected lncRNAS were distributed among mouse chromosomes 1–19, X and Y, and lncRNAS had the largest number on the second chromosome (2,925, 10%) (Figure 2C).

We detected 30,450 lncRNAs in the hippocampus of the mouse by RNA-seq. Estimate the expression levels of various RNAs between the two groups based on the FPKM value. Compared with the control group, there were 229 lncRNAs siginificantly altered in HE mice (113 upregulated and 116 downregulated) (Supplementary Table S2). 49 known miRNAs were significantly altered (37 upregulated and 12 downregulated) (Supplementary Table S3). In addition, 282 mRNAs were upregulated in the HE group, while 81 were downregulated (Supplementary Table S4). The differential expression of lncRNAs, miRNAs, and mRNAs between the two groups was visually displayed by volcano and heat maps (Figure 3). In HE mice, the number of upregulated miRNAs and mRNAs were greater than that downregulated miRNAs and mRNAs, while the number of upregulated and downregulated lncRNAs were similar.

Quantitative real-time PCR (qRT-PCR) was used to confirm the accuracy and reproducibility obtained from RNA-seq analysis. We randomly selected four mRNAs (CLDN2, AQP1, CRHR2 and EPN3), four miRNAs (miR-1264-5p, miR-34b-5p, miR-376c-5p and miR-743b-3p) and three lncRNAs (ENSMUST00000124806, ENSMUST00000227933 and ENSMUST00000177220) for qRT-PCR analysis. The qRT-PCR results were consistent with the trend of RNA-seq data (Figure 4), which indicated that the RNA-seq results were reliable (Considering that we performed RNA-seq with three mice per group, this only met the minimum number of statistical analyses. Therefore, it is recommended that researchers verify the expression status of the genes to be studied before conducting experiments).

GO analysis was performed to elucidate the roles of differentially expressed mRNAs identified by RNA-seq. The results showed that differentially expressed mRNAs were significantly enriched in 51 GO entries. Upregulated mRNAs were primarily enriched in cellular process, single-organism process, signaling and behavior in biological processes; most of the cellular components were enriched in cell, synapse, organelle and membrane; molecular functions were mostly enriched in binding, catalytic activity and transporter activity (Supplementary Figure S2A). On the other hand, the downregulated mRNAs were primarily enriched in positive regulation of biological process, regulation of biological process; cell components were mostly enriched in cell, synapse part and macromolecular complex; molecular functions were mostly enriched in binding, nucleic acid binding transcription factor activity (Supplementary Figure S2B; Supplementary Table S5).

KEGG pathway analysis revealed that there were 117 pathways enriched in upregulated mRNAs and 23 enriched in downregulated mRNAs (Supplementary Table S6). Neuroactive ligand-receptor interaction, ECM-receptor interactions, Tight junction, Huntington disease, Glycine, serine and threonine metabolism, Cell adhesion molecules (CAMs), AMPK signaling pathway were the most enriched, indicating that dysregulated mRNAs may play important roles in nervous system dysfunction. The top 20 pathways of mRNAs are shown in Figures 5A,B (Supplementary Table S7).

We built interaction networks of lncRNA-miRNA-mRNA based on the expression profiles in HE mice. First, lncRNAs and mRNAs targeted by differentially expressed miRNAs were predicted, respectively. A total of 73 lncRNAs, 39 miRNAs and 134 mRNAs interactions were identified (Supplementary Table S8). We constructed co-expressed lncRNA-miRNA-mRNA visualization networks using cytoscape (v3.6.0) (Figure 6). In the networks,mmu-miR-7667-5p (degree = 33), miR-490-y (degree = 28) and mmu-miR-743b-3p (degree = 24) had more target mRNAs. In addition, mmu-miR-3064-5p (degrees = 14), miR-285-z (degrees = 12) and mmu-miR-34b-5p (degrees = 12) had more target lncRNAs. The results showed that a single mRNA or lncRNA could be correlated with one or more miRNAs and vice versa. Therefore, the interaction between lncRNAs, miRNAs and mRNAs may mediate the progression of HE.

Next, mRNAs related to the nervous system and its associated lncRNAs, miRNAs were selected to produce ceRNA networks (Figure 7; Supplementary Table S9). The sub-network contained five core miRNA nodes (miR-743b-3p, miR-376c-5p, miR-708-y, mmu-miR-1264-5p, miR-34b-5p). It was predicted that Jade2, Fos and Npas4 were the target genes of miR-34b-5p. Aqp1 was the target gene of miR-743b-3p. According to KEGG analysis, these mRNAs were key genes for nervous system development, synaptic transmission, synaptic plasticity, and synapse assembly. The above results indicate that these lncRNAs might interact with miRNAs to regulate the expression of mRNAs’ expression and play important roles in nervous system development, synaptic transmission, and synaptic organization of HE mice.

In the ceRNA network, SIX3OS1 was one of the significantly upregulated lncRNAs. We constructed its shRNA lentiviral vectors and transduced them into cortex neurons in vitro to assess the roles of SIX3OS1 (The number of primary cultured hippocampal neurons is limited, and it is difficult to obtain enough samples for detection. Therefore, cortical neurons were used to verify the regulatory relationship between SIX3OS1 and its predicted targeted genes). SIX3OS1 silencing was confirmed by RT-PCR (Figure 8A). RT-PCR analysis of SIX3OS1-shRNA cells showed that the expression of miR-743b-3p targeted by SIX3OS1 was upregulated; AQP1, EBF2, NKAIN3 and ISL1 were downregulated (Figures 8B–F). We further explored the effects of SIX3OS1 on the dendritic spines and synaptic transmission of NH₄Cl-treated neurons. The results showed that dendritic spine density and maturity of the NH₄Cl and NH₄Cl + sh-NC groups significantly decreased compared with the control group. The NH₄Cl + sh-SIX3OS1 group was similar to the control group (Figures 9A–C).

Studies have shown that the imbalance of excitatory/inhibitory neurotransmitters is one of the major causes of HE (Rao, 2002; Nardone et al., 2016). Therefore, we measured the inhibitory and excitatory transmission of neurons transfected with sh-SIX3OS1 during hyperammonemia by whole-cell patch-clamp recordings. Compared with the control group, hyperammonemia significantly increased the amplitude and frequency of hippocampal neuron mIPSC, and decreased the amplitude and frequency of mEPSC. This change could be reversed after transfection with sh-SIX3OS1(Figures 9D–I). The above research showed that SIX3OS1 was essential for the maintenance of synaptic function.