In this study, a total of 44 peripheral blood samples were collected from patients with AIH at the West China Hospital of Sichuan University. The diagnostic criteria adhered to the International Autoimmune Hepatitis Group (1999) guidelines (Alvarez et al., 1999; Yeoman et al., 2009). The clinical characteristics of these patients are listed in Table 1. The histological features of patients are shown in Supplementary Figure S1 and Supplementary Table S1. 36 healthy subjects were studied as controls. Among the samples, the peripheral blood samples from 27 patients and 21 normal subjects were used to analyze the proportion of mature DCs in the peripheral blood by flow analysis, and blood samples from 17 patients and 15 HCs were used to isolate moDCs for the experiments described below. The study was approved by the Independent Ethics Committee of West China Hospital and conducted in accordance with the relevant principles.

Differentiation of human monocyte into moDCs was performed as previously described (Wang et al., 2016). Human lymphocyte separation medium (Dakewei, Shenzhen, China) was used to isolate peripheral blood mononuclear cells (PBMCs) from whole blood samples collected from AIH patients and HCs. CD14⁺ monocytes were separated from PBMCs using CD14⁺ magnetic beads (Miltenyi Biotec, Gladbach, Germany). After magnetic bead sorting, the proportion of CD14⁺ cells were as high as 94% (Supplementary Figures S2A,B). Sorted cells were then cultured for 5–7 days in medium containing 50 ng/ml granulocyte/macrophage colony-stimulating factor (GM-CSF; Novoprotein, Shanghai, China) and 20 ng/ml IL-4 (Novoprotein, Shanghai, China) at a concentration of 1.5 × 10⁶ cells/ml. Then, 0.5 μg/ml lipopolysaccharide (LPS; Sigma–Aldrich, United States) was added to the medium on day 5, and the cells were cultured for another 1–2 days to become mature moDCs. Flow cytometry was used to identify moDCs, and the moDCs displayed high expression of CD11c (APC, BioLegend, United States) (Supplementary Figures S2C,D).

PBMCs collected from the peripheral blood of volunteers and moDCs obtained after 24 h of LPS stimulation were washed twice with phosphate-buffered saline (PBS, Servicebio, Wuhan, China). Then, approximately 1 × 10⁶ prepared cells were suspended in 100 μl of PBS and stained with different kinds of fluorochrome-coupled antibodies for 30 min at 4°C. Data were acquired using a CytoFLEX flow cytometer (Beckman Coulter, Life Science, United States) or NovoCyte (ACEA, Hangzhou, China), and the percentages of Lineage (CD3/14/16/19/20/56; BioLegend, United States), CD11c⁺(APC, BioLegend, United States), CD80⁺ (FITC, BioLegend, United States), CD86⁺ (PerCP-Cy5.5, BioLegend, United States) and HLA-DR⁺ (PE, BioLegend, United States) DCs were analyzed. And all these antibody isotypes including APC-IgG1, PerCP-cy5.5-IgG2b, FITC-IgG1, PE-IgG2a and FITC-IgG2b (BioLegend, United States) were used in the analysis (Supplementary Figure S3).

Serum from AIH patients and healthy controls was collected by centrifugation of whole peripheral blood at 1,000 g for 10 min. The levels of IFN-γ, TNF-α and IL-12 were measured using human ELISA detection kits according to the manufacturer’s recommendations (MultiSciences, Hangzhou, China). The final density values were measured at 450 and 570 nm by a microplate reader (BioTek, Winooski, VT, United States).

Total RNA was extracted from moDCs isolated from patients (called group A) or HCs (called group H) using TRIzol reagent (Tiangen, Beijing). An Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, United States) and Qubit Fluorometer (Invitrogen) were used to assess RNA quality. Samples with an RNA integrity number (RIN) > 7.0 and a 28S:18S ratio > 1.8 were used in subsequent experiments.

A total amount of 3 µg of total RNA per sample was used for the construction of a small RNA library. Following the manufacturer’s recommendations, sequencing libraries were prepared by using an Illumina TruSeq stranded total RNA with Ribo-Zero gold kit, and index codes were added to attribute sequences to each sample. The libraries were sequenced on a HiSeq X Ten platform (Illumina, San Diego, CA, United States). Similarly, a miRNA library was constructed from samples with Illumina’s TruSeq small RNA library preparation kit. The libraries were sequenced on an Illumina HiSeq 2500 sequencing platform. The whole procedures were performed by CapitalBio Technology (Beijing, China).

A paired t test and the Benjamini-Hochberg method were used to detect DE mRNAs and DE ncRNAs in group A versus group H. The thresholds for significantly DE mRNAs, DE miRNAs, DE lncRNAs and DE circRNAs were |log2FC|≥1 and p < 0.05. Novel lncRNA, named MERGE, is an assembled transcript, which removes transcripts that are smaller than 200 bp and predicted by the retention software CPC, CNCI and HMMER that there is no coding ability. miRDeep2 is used for the prediction of novel miRNAs. circRNA identification uses STAR for fusion comparison, CIRCexplored2 for back-splicing comparison, and finally determines as circRNA. Heatmaps and volcano plots were drawn to describe the DE mRNAs and DE ncRNAs.

Based on the raw data for DE lncRNAs, DE circRNAs, and DE mRNAs, we calculated the Pearson correlation coefficients of each DE mRNA and DE lncRNA/circRNA. Significant mRNA-lncRNA pairs and mRNA-circRNA pairs were detected by a correlation test with the thresholds of p < 0.05 and r < −0.8. The mRNA-lncRNA/circRNA coexpression relationships were then defined.

GO analysis is a functional analysis related to DE mRNAs based on the GO categories (http://www.geneontology.org). The GO categories consist of three structured networks of defined terms named cellular component (CC), molecular function (MF), and biological process (BP). Based on the latest KEGG (https://www.genome.jp/kegg) database, KEGG pathway analysis was used to detect the biological pathways related to DE mRNAs.

ceRNA networks are used as decoys for miRNA binding, which can counteract the gene silencing activity of a miRNA. Based on the principles for ceRNA network construction, we integrated the coexpression relationships of lncRNAs and mRNAs and those of miRNAs and circRNAs and focused more on the mRNAs regulated by miRNAs and lncRNAs with a positive relationship. The difference with fold change (FC) ≥ 2 or ≤ 0.05 and p < 0.05 was considered statistically significant and the FDR was calculated to correct the p value on RNA-sequencing analysis. Besides there are other additional conditions we used for screening: the average expression level of at least one group in AIH/HC is more than 0.5; the number of expressed samples accounts for 2/3 of the total number of samples. Cytoscape was used to construct the lncRNA-miRNA-mRNA ceRNA networks. Similarly, we detected the miRNAs that could simultaneously regulate circRNAs and mRNAs and analyzed the coexpression relationships between these circRNAs and mRNAs. The circRNA-miRNA-mRNA ceRNA networks were constructed with Cytoscape.

Quantitative RT-qPCR was used to validate the different expression levels of ncRNAs in moDCs from AIH patients and healthy controls. Total RNA was extracted from moDCs using TRIzol reagent (Ambion, Thermo Fisher Scientific, United States). cDNA was synthesized from 1 μg of extracted total RNA with the PrimeScrip RT reagent kit (Takara, Shiga, Japan) and amplified by real-time qPCR with SYBR Green Supermix on a CFX96 RT-qPCR detection system (BioRad, Hercules, CA, United States). The expression levels of ncRNAs were normalized to actin expression level. All primers were obtained from Tsingke (Beijing, China).

Statistical analysis was performed with SPSS 22.0 software and GraphPad Prims 9. A two-tailed Student’s t-test was used to detect significant differences between groups. The results are expressed as the mean ± SD, and significance was defined at p < 0.05. The expression level of each ncRNA was represented as 2^−Δct values normalized to actin expression on real-time qPCR analysis. p values < 0.05 were considered statistically significant.

Blood samples from 27 AIH patients and 21 HCs were studied and the clinical characteristics of these patients are listed in Supplementary Table S2. We analyzed the subtypes of DCs in the peripheral blood of the patients and HCs. The percentages of CD11c⁺, CD86⁺ and HLA-DR⁺ DCs between the two groups were compared. The frequency of CD11c⁺CD86⁺ DCs was significantly higher in the patients with AIH than in the HCs. The proportion of CD11c⁺HLA-DR⁺ DCs exhibited an upward trend in the AIH patient group but without statistical significance (Figure 1A). To validate the important role of mature DCs, we performed mature DC-related cytokine analysis to assess human IFN-γ and TNF-α in the peripheral blood of patients with AIH and HCs. The levels of IFN-γ were significantly increased in the AIH group compared to the HC group, while there was no significant difference in TNF-α between the two groups (Figure 1B). Overall, circulating mature DCs may be involved in the pathogenesis of AIH. We randomly selected five patients with AIH and analyzed the proportion of mature DCs in the peripheral blood again after 2–3 weeks of meprednisone (MP) treatment. The clinical characteristics of these patients are listed in Supplementary Table S3. The level of IgG in the five patients decreased after treatment (p = 0.03), and total bilirubin (TBil), ALT, and AST levels showed downward trends but without significant differences (Supplementary Figure S4). In addition, we found that the proportion of mature DCs in the peripheral blood of AIH patients after treatment was lower than that before treatment. The frequency of CD11c⁺HLA-DR⁺ DCs was significantly lower after treatment than before treatment. Furthermore, the percentage of CD11c⁺CD86⁺ DCs also exhibited a downward trend but without statistical significance (Figure 1C). These results indicated that DCs participate in regulating the occurrence and development of AIH in patients.

We isolated moDCs from the peripheral blood of AIH patients and HCs and used GM-CSF, IL-4 and LPS to induce mature moDCs. We evaluated the morphology, immunophenotypic characteristics, and cytokines of mature moDCs from AIH patients compared with those of mature moDCs from HCs. As shown in Figure 2A, the AIH patients had more mature DCs. The mean percentage of CD11c⁺CD80⁺ DCs in the AIH patient group was significantly higher than that in the HC group (1.37 ± 0.75% vs. 19.80 ± 8.42%, p < 0.01). The mean proportion of CD11c⁺CD86⁺ DCs in the HC group was 21.93 ± 4.60%, which was lower (p < 0.05) than that in the AIH group (59.92 ± 3.40%). The mean percentage of CD11c⁺HLA-DR⁺ DCs in the AIH group was 60.51 ± 3.17%, which was higher (p < 0.05) than that in the HC group (28.4 ± 7.34%) (Figure 2B). We compared the supernatant levels of cytokines related to mature DCs between the AIH group and HC group. The levels of IFN-γ were significantly increased in AIH patients. The levels of IL-12 exhibited an upward trend but without statistical significance (Figure 2C).

According to whole RNA-seq data, we analyzed DE ncRNAs (lncRNAs, circRNAs, miRNAs) and mRNAs with |log2FC| ≥ 1 and p < 0.05. The DE ncRNAs identified between the two groups are shown in a heatmap (Figure 3A) and volcano plot (Figure 3B). In total, there were 352 DE lncRNAs (178 upregulated and 174 downregulated), 143 DE circRNAs (32 upregulated and 111 downregulated), 33 DE miRNAs (20 upregulated and 13 downregulated), and 550 DE mRNAs (292 upregulated and 258 downregulated) in the moDCs of patients with AIH compared to those of HCs. In addition, the most upregulated lncRNA, circRNA, miRNA and mRNA were MERGE.24561.12, hsa_circ_0000279, hsa-miR-7844-5p, and AP000295.1, respectively. The most downregulated lncRNA, circRNA, miRNA and mRNA were MERGE.3901.1, hsa_circ_0006719, hsa-miR-889-3p, and ST13P19, respectively (Table 2).

GO analysis was conducted to identify the mRNAs involved in lncRNA-miRNA-mRNA and circRNA-miRNA-mRNA networks. As shown in Figure 4A, GO analysis divided these DE mRNAs into three terms: biological process, cellular component, and molecular function. In the lncRNA-miRNA-mRNA network, the DE mRNAs were mainly involved in protein binding (molecular function), the intracellular part (cellular component), and the regulation of a protein metabolic process (biological process). In the circRNA-miRNA-mRNA network, the DE mRNAs were mainly involved in the intracellular organelle part (cellular component), a macromolecule metabolic process (biological process), and protein binding (molecular function). We performed KEGG pathway analysis to detect the key pathways and relationships between the mRNAs and ncRNAs. KEGG pathway analysis revealed the top 30 pathways enriched in the DE mRNAs related to the coexpression network. In the lncRNA-miRNA-mRNA networks, signal transduction, MAPK signaling pathway, Wnt pathway and the immune system were the most enriched pathways. In the circRNA-miRNA-mRNA networks, cytokine signaling pathway, metabolism, and the immune system were the most enriched pathways (Figure 4B). Tables 3, 4 list the top 10 coexpressed pairs of lncRNAs/circRNAs and mRNAs in moDCs. Then, we combined the results of the miRNA-lncRNA/circRNA joint analysis and miRNA-mRNA joint analysis. Due to the complicated and huge ceRNA networks, we used more stringent parameters to reduce the number of networks. We illustrated the ceRNA networks based on the key DE mRNAs including CIITA, PDXK, LRP1, PTPN7, RBM39, OAS2, WDR19, IRGQ, ST8SIA4, OS9, SKP1, OXNAD1, SMARCB1, BUD23, GPNMB, TNIP2, GPB3 and PIK3R2. Then 24 lncRNA-miRNA-mRNA pathways were constructed, including 16 lncRNAs, 11 miRNAs and 16 mRNAs. Five circRNA-miRNA-mRNA pathways were constructed, including three circRNAs, two miRNAs and three mRNAs (Figure 4C).

To validate our RNA-seq data, we selected five DE lncRNAs and DE circRNAs respectively and enlarged the sample sizes to measure the expression levels of these ncRNAs among 17 AIH patients and 15 healthy controls by using qRT-PCR. The selection criteria of these ncRNAs were as followings: The difference with fold change (FC) ≥ 2 or ≤ 0.05 and p < 0.05 was considered statistically significant. Besides the average expression level of at least one group in AIH/HC is more than 0.5 and the number of expressed samples accounts for 2/3 of the total number of samples. The clinical characteristics of these patients are listed in Supplementary Table S4 and the primers we used are listed in Table 5. The experimental results indicated that the expression levels of ENST00000543334 and hsa_circ_0000279 were upregulated, while those of hsa_circ_0005076 were downregulated, which was consistent with the RNA-seq data. In addition, the expression levels of ENST00000445461, ENST00000523995, hsa_circ_0004058, and hsa_circ_0004853 showed the same trends as the RNA-seq results but without significance (Figure 5).