MicroRNA-148a facilitates inflammatory dendritic cell differentiation and autoimmunity by targeting MAFB

DCs comprise a heterogeneous family of professional antigen-presenting cells that maintain the homeostasis of the immune system (1–4). DC populations in the skin have been well investigated in mice (5–7). In the steady state, two major subsets of conventional DCs (cDCs) have been characterized in the mouse nonlymphoid tissues, CD103⁺CD11b^– cDCs and CD103^–CD11b⁺ cDCs, whereas the epidermis contains Langerhans cells. In the inflammatory state, peripheral monocytes are rapidly recruited to inflammatory sites and give rise to monocyte-derived DCs (moDCs) (5–7). Moreover, an increased frequency and activation status of plasmacytoid DCs (pDCs), a potent type I IFN-producing DC subset, have been documented in early psoriatic lesions (8, 9). Both cDCs and pDCs originate from DC precursors in a Fms-related tyrosine kinase 3 ligand (Flt3L)–dependent manner, whereas the generation of moDCs is GM-CSF dependent (1, 2).

Up to now, the regulation of moDC activation in autoimmune diseases has been studied extensively (10–13). Studies have demonstrated that transcriptional factors, such as STAT5, PU.1, and MAFB, affect moDC generation at different stages of hematopoietic differentiation (14–20). Both PU.1 and MAFB are involved in the myeloid commitment of hematopoietic stem cells, and the balance between them specifies alternative DC or macrophage fate (18, 19). Despite intensive study, the molecular mechanism underlying regulatory network for moDC generation is far from clarified.

Psoriasis is a common, chronic autoinflammatory skin disease characterized by demarcated, red, and scaly plaques. Environmental and genetic factors trigger the hyperproliferation and disturbed differentiation of keratinocytes, thickening of the epidermis, formation of new blood vessels, and accumulation of leukocytes in epidermis and dermis, of which T cells and DCs are the most critical (21, 22). Psoriatic skin contains large numbers of activated DCs, which produce IL-23, IL-12, IL-6, and TNF-α. They activate and polarize autoaggressive Th cells toward Th17, Th1, and Th22 cell subsets defined by the production of IL-17A, IL-17F, IFN-γ, and IL-22, respectively (9, 23, 24). Recently, the involvement of different skin DC subsets in regulating psoriatic plaque formation has been studied extensively. pDCs and cDCs are dispensable for the local skin inflammation in psoriasis, since E2-2^–/– mice and Flt3L^–/– mice presented no significant difference in ear swelling and epidermal thickening compared with WT mice (25, 26). The Ccr2^–/– mice, which have profound monocytopenia due to a defect in monocyte egress from BM, showed a significant reduction in skin pathological phenotype (26). Depletion of monocytes in vivo through the use of antibody also greatly relieved the symptoms (26). The function of Langerhans cells is controversial because Lan-DTR mice developed a degree and course of psoriasiform skin disease similar to those of WT mice in an imiquimod-induced (IMQ-induced) model (25) but showed a certain degree of improvement in an IL-23–induced model (26). The finding that locally increased expression of GM-CSF and IL-23 was exclusively produced by Langerin^– DCs in vivo, which further supports moDCs as being the most important DC subset in psoriasis (25).

miRNAs are short (~22 nt), evolutionarily conserved, single-stranded RNAs that control the expression of complementary target mRNAs, leading to their transcript destabilization, translational inhibition, or both (27). miRNAs are critical regulators of immune cell development and function (28). In this study, we used the psoriasis disease model to identify miR-148a as a paramount regulator for autoimmune-related moDC differentiation. miR-148a deficiency led to decreased pathogenic moDCs. Consequently, either miR-148a knockout or miR-148a inhibition by intradermal administration with antagomir-148a prevented the development of moDCs and psoriasis-like inflammation in the IMQ-induced psoriasis-like mouse model. The mechanism research revealed that MAFB was a bona fide direct target of miR-148a, which was transcriptionally regulated by PU.1. Therefore, these results identify a pivotal role for the PU.1-miR-148a-MAFB pathway in the differentiation of moDCs and suggest a potential therapeutic avenue for inflammatory moDCs mediated autoimmune diseases.

miR-148a is indispensable for moDC differentiation. Proinflammatory moDC infiltration is critical for the progression of several autoimmune diseases, including systemic lupus erythematosus (SLE), psoriasis, inflammatory bowel disease, and so on (25, 29, 30). A comprehensive analysis of our previous data in patients with SLE (31) or psoriasis revealed an increase in the expression of miR-148a-3p (referred to as miR-148a here) (Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.133721DS1). Other groups also reported similar elevation in autoimmune diseases (32). Therefore, we hypothesized that miR-148a might affect the development or function of pathogenic moDCs in autoimmune diseases.

Before evaluating the function of miR-148a in psoriatic inflammation, we first checked the effect on hematopoietic differentiation in the steady state (Supplemental Figure 2, A–D). Accordingly, different cell subsets in the spleen and skin from miR-148a^–/– and WT littermate mice were detected (Supplemental Figure 2, A and C). Consequently, there was no difference in the number of lymphoid or myeloid cells (Supplemental Figure 2, B and D), indicating that miR-148a was dispensable for hematopoietic differentiation in the steady state.

In the inflammatory state, Ly6C^hi monocytes migrate to the lesion sites and then differentiate into moDCs. To validate whether the DC differentiation from monocytes was affected by miR-148a, Ly6C^hi monocytes were isolated from BM of miR-148a^–/– and WT mice and then cultured with GM-CSF and IL-4. It has been demonstrated that Ly6C^hi monocytes can give rise to both CD11c⁺MHCII^hiCD11b^int DC and CD11c⁺MHCII^intCD11b^hi macrophages (10). The conclusion was confirmed in the culture system that sorted CD11c⁺MHCII^hiCD11b^int cells exhibited DC morphology and CD11c⁺MHCII^intCD11b^hi cells had a typical macrophage morphology (Figure 1, A and B), allowing us to evaluate the differentiation potential of monocytes toward macrophages and DCs in the same culture system.

Figure 1

miR-148a is indispensable for moDC differentiation. (A and B) Monocytes isolated from BM of WT or miR-148a^–/– mice were cultured with 50 ng/mL GM-CSF and 20 ng/mL IL-4 for 5 days. Flow cytometry analysis (A), morphology (B), and statistical data (C) of moDCs and macrophages are shown. (D) Flow cytometry and statistical data of apoptosis during monocyte differentiation at indicated time points were analyzed (n = 3). Scale bars: 10 μm (B). P values were determined by 2-tailed unpaired t test. Data are shown as mean ± SEM. (***P < 0.001, n.s., not significant). miR-148a, microRNA-148a; moDC, monocyte-derived DC.

There was no difference of the expanding potential of monocytes between control and miR-148a deficiency, since the number of live cells was similar (Figure 1C). The monocytes of the miR-148a^–/– group had considerably less moDCs (about 70% reduction) than those of controls (Figure 1, A and C). In contrast, both the percentage and the number of macrophages increased in the miR-148a^–/– group (Figure 1, A and C), indicating a critical role for miR-148a in promoting the differentiation of monocytes toward inflammatory DCs at the expense of macrophage.

Previous studies reported that miR-148a was involved in antiapoptosis and frequently upregulated in primary and tumor cells (33, 34). To confirm whether the decrease in the proportion and number of miR-148a^–/– moDCs was a result of the increased apoptosis, the entire cells were stained with annexin V and propidium iodide. Consequently, the examination of cell apoptosis showed no difference between miR-148a^–/– and WT throughout the inducing process, suggesting that miR-148a did not affect the moDC survival (Figure 1D).

The transcriptional regulation of moDC differentiation has been extensively studied (2, 13). The balance of MAFB and PU.1 was considered to specify alternative macrophage or DC fate (18). The differentiation of monocytes to DCs results in a rapid increase of PU.1, achieving high levels that precede the phenotypic changes, while MAFB expression is decreased during the differentiation. Inhibited MAFB expression in monocytes appears to be required for DC specification, since constitutive MAFB expression inhibits DC differentiation (18). In addition, MAFB deficiency is found to specifically enhance sensitivity to M-CSF and cause activation of the myeloid master regulator PU.1 in hematopoietic stem cells in vivo (19). One possible mechanism for the PU.1 regulation of MAFB function is that PU.1 can directly bind to MAFB and inhibit MAFB transactivation capacity. In this research, we found that miR-148a mediated the inhibition of MAFB expression through PU.1 regulation, suggesting a new mechanism for regulating the balance of MAFB and PU.1 and, thereby, the differentiation of moDCs. Interestingly, a recent study reported that PU.1 could directly bind to the superenhancer of miR-148a and facilitate its maturation in mouse primary B cells (36). Whether this regulation exists in primary monocytes and moDCs remains to be made clear.

miRNA is generally regarded as a multifunctional molecule and can exert diverse roles in different immune cells by targeting different molecules intracellularly (28). Increased expression of miR-148a, which occurs frequently in patients with lupus and lupus-prone mice, impaired B cell tolerance — through promoting the survival of immature B cells upon B cell receptor engagement via suppressing the expression of GADD45a, PTEN, and BCL2l11 — and facilitated the development of lethal autoimmune disease (33). miR-148a could also considerably reduce the DNMT1 protein level, which was one of the major epigenetic components that linked to DNA hypomethylation in T cells in SLE (31). There was no significant difference in the number of B cell and T cell subsets between miR-148a^–/– mice and WT mice in the steady state. Hence, miR-148a might affect the survival and function of both autoreactive B and T cells in the inflammatory state. To confirm the cell-intrinsic role of miR-148a in moDCs, we designed two experiments. First, we constructed BM chimeric mice and found that miR-148a^–/– chimeric mice demonstrated less psoriatic skin inflammation, indicating that miR-148a takes effect only on immune cells, rather than other skin tissue cells during the process. Second, monocytes were depleted in vivo through the use of antibodies before and during IMQ painting. Monocyte depletion greatly alleviated the severity of the psoriasiform skin inflammation. In addition, there was no significant difference in the symptoms between miR-148a^–/– mice and WT mice, indicating that miR-148a in moDCs is more likely to be critical for the IMQ-induced psoriasis-like inflammation.

It has also been reported that miR-148a regulates the innate immune response. It has been demonstrated that miR-148a negatively regulates the TLR4-triggered functional maturation of moDCs by targeting CaMKIIa (34), and inhibits macrophage function by directly targeting several upstream regulators of NF-κB and STAT3 signaling, including GP130, IKKα, IKKβ, IL1R1, and TNFR2 (37). We also found that miR-148a^–/– moDCs could produce increased levels of TNF-α and IL-6 in response to LPS stimulation (data not shown). However, a less than 10% increase was rather limited compared to the decreased potential of moDC differentiation both in vitro (about 70% reduction) and in vivo (a more than 1-fold reduction), as shown in Figure 1C, Figure 4I, and Figure 6H. Therefore, miR-148a in monocytes exerted deleterious effects during psoriasis.

In addition to miR-148a, several other miRNAs have been reported to regulate the differentiation and function of moDCs, such as miR-21, miR-34a, and miR-155 (38–41). Besides miR-148a, both miR-21 and miR-34a also affect moDC generation, and miR-155 promotes the apoptosis and activation of moDCs. As molecule fine-tuning gene expression, these miRNA targets generate a complex regulatory network that comprise different levels, including miR-21/WNT1 and miR-34a/JAG1 at the ligand/receptor level, miR-155/SOCS1 at the signal transduction level, and miR-148a/MAFB at the transcriptional factor level. Interestingly, miR-21 was found to be abnormally elevated in psoriasis, and the inhibition has been shown to have a beneficial effect on the treatment of psoriasis (42). In addition, miR-155 is also upregulated in psoriatic biopsy samples and responsible for maintaining the inflammation in psoriasis (43).

In conclusion, the present study provides evidence that miR-148a, which could be transcriptionally activated by PU.1, promotes moDC differentiation through directly targeting MAFB, thereby causing the immune imbalance and inflammatory response in skin lesions of psoriasis (Figure 8). Since inflammatory DCs are pathogenic in various autoimmune diseases (SLE, RA, etc.), this research suggests that miR-148a could serve as a potential target for the treatment of not only psoriasis but also other autoimmune diseases.

Figure 8

Regulation model. miR-148a, which could be transcriptionally activated by PU.1, promotes moDC differ-entiation by directly targeting MAFB, thereby causing the immune imbalance and inflammatory response. miR-148a, microRNA-148a; moDC, monocyte-derived DC.

Human subjects. A total of 34 psoriatic patients, who were diagnosed with psoriasis vulgaris, were recruited from outpatient clinics at Huashan Hospital of Fudan University. Psoriasis disease activity was assessed using PASI score. The patient information is shown in Supplemental Table 1. In addition, a total of 51 sex- and age-matched healthy controls were recruited from Renji Hospital of SJTUSM.

Mice. The C57BL/6J mice were acquired from Slack Company. miR-148a-knockout mice B6/JNju-Mir148a^em2Cd262in1/Nju on a [C57BL/6] background were provided by Model Animal Research Center, Nanjing University. In addition, male mice at 6–8 weeks of age were used to carry out all the following experiments. All mice were bred and housed under the conditions of specific pathogen–free circumstance.

IMQ-induced psoriasis-like mouse model and antagomir-148a and MAFB siRNA treatment. Male miR-148a^–/– mice on a C57BL/6J background (6–8 weeks of age) were daily painted with IMQ cream (5%) (Aldara; 3M Pharmaceuticals) on the shaved back for the indicated days and then sacrificed for the subsequent experiments. The control mice were applied with the same dose of vehicle cream. Male miR-148a^–/– mice and their littermates (6–8 weeks of age) were injected intradermally with antagomir-148a (10 nM in 150 μL PBS) or antagomir-NC (10 nM in 150 μL PBS) on days 0, 1, 2, and 3 during the application of IMQ (44). Antagomir-148a (GenePharma, B05001) and the corresponding negative control antagomir-NC (GenePharma, B04007) were purchased from Shanghai GenePharma Co. Ltd. As for MAFB siRNA treatment, male miR-148a^–/– mice and their littermates (6–8 weeks of age) were intradermally administered siRNA specific for MAFB (10 nmol in 100 μL PBS) or NC (10 nmol in 100 μL PBS) on days 0, 1, 2, and 3 during the application of IMQ. In addition, MAFB siRNA (Ambion, 4457308) and NC siRNA (Ambion, 4457289) were purchased from Applied Biosystems.

Scoring severity of skin inflammation in psoriasis-like mouse model. An objective scoring system was developed based on the clinical PASI for assessing the patients with psoriasis. Accordingly, erythema, thickening, and scaling were each scored according to the scope (0–4: 0, none; 1, slight; 2, moderate; 3, marked; 4, very marked). The overall score was obtained by calculating a sum of the 3 index scores (score 0–12).

Cell isolation, flow cytometry analysis, and cell culture. Skin cells were prepared on the basis of previous research with minor modifications (45). In brief, the epidermis and dermis were separated by adopting the dispase (25 U/ml in Dulbecco’s DPBS) and then the dermal cell suspension was acquired using collagenase and hyaluronidase. Splenocytes were prepared by digesting the tissue fragments with collagenase type IV and DNAse I. Antibodies used for cell staining were presented in Supplemental Table 2. Cells were analyzed using Fortessa (BD Biosciences) and FlowJo software.

PBMCs were separated from the peripheral blood of healthy controls and patients with psoriasis by density gradient centrifugation (GE Healthcare). Human and mouse monocytes were isolated with the Human Monocyte Isolation Kit II and Mouse Monocyte Isolation Kit (BM) (Miltenyi Biotec), respectively. Then, the cells were cultured in RPMI 1640 medium (Gibco) with 10% FBS (Gibco) at 37°C and 5% CO₂, or collected for subsequent experiments. HEK 293T cells and HeLa cells were maintained in DMEM (Gibco) supplemented with 10% FBS (Gibco). To obtain moDCs, BM cells or Ly6C^hi monocytes separated from BM were cultured with 100 ng/mL GM-CSF and 100 ng/mL IL-4 (Peprotech) for indicated days. The Annexin V Apoptosis Detection Kit (BioLegend) was used for apoptosis assays according to the manufacturer’s protocol.

siRNAs and agomir transfection. BM cells, mouse monocytes isolated from BM, or human monocytes isolated from peripheral blood were transfected with Lipofectamine RNAiMAX (Invitrogen) according to the protocol. Hela cells and HEK 239T cells were transfected with Lipofectamine 3000 (Invitrogen). Briefly, small interfering RNA (siRNA; 200 nM), agomir (200 nM), plasmids, and transfection reagents were diluted with Opti-MEM (Gibco) and then gently mixed. After incubation for 15 minutes at room temperature, the transfection mixture was added to the cell culture. Cells were harvested for subsequent experiments. The agomir-148a (GenePharma, B06002) and the corresponding negative control (GenePharma, B04008) were purchased from Shanghai GenePharma Co. Ltd. Human PU.1 siRNA (Santa Cruz, sc-36330), mouse PU.1 siRNA (Santa Cruz, sc-36331), and MAFB siRNA (Santa Cruz, sc-35840) were purchased from Santa Cruz Biotechnology.

Histology. As for H&E staining, the histological analysis was conducted as Lai et al. reported (46). First, mouse skin tissues were fixed in formalin and embedded in paraffin and then stained with H&E. Finally, to assess dermal thickness, we measured the thickness from the stratum papillare to the stratum reticulare and the epidermal hyperplasia from the basal layer to the stratum corneum,using Case Viewer software (3DHISTECH Ltd). The measurement of skin thickness was conducted on 6 randomly selected areas per ×10 field from 3–5 fields per section. The obtained skin thickness was first averaged per mouse and then used to calculate the average per treatment group for the following statistical analysis. As for immunochemistry, the mouse skin paraffin sections were stained with hematoxylin and mouse monoclonal anti-CD11c antibody (GB11059, Servicebio).

Construction of BM chimeras. B6.CD45.1 recipients were sublethally irradiated by X-ray (8 Gy) and then intravenously transferred with 5 × 10⁶ BM cells derived from miR-148a^–/– mice or C57BL/6J mice. Chimeras were adopted for experiments 6 weeks after the initial reconstitution.

Depletion of neutrophils and monocytes. The depletion in vivo was conducted based on the report by Singh et al. (26). In brief, as for the depletion of neutrophils alone, the mice were intraperitoneally administered 500 mg antibody anti-Ly6G (clone 1A8, BioXCell) on days 0, 1, 2, 3, and 4. Meanwhile, to deplete monocytes and neutrophils, the mice were injected intraperitoneally with 500 mg of antibody anti-Gr1 (clone RB6-8C5, BioXCell) at the same time points as above.

mRNA-Seq and miRNA-Seq. All the RNA was first extracted by using the miRNeasy Micro Kit (QIAGEN) and determined with Bioanalyzer 4200 (Agilent). Then, the preparation for the next-generation mRNA libraries was performed with VAHTS mRNA-Seq v2 Library Prep Kit for Illumina (Vazyme), whereas the preparation for the miRNA libraries was performed with NEBNext Multiplex Small RNA Library Prep Set for Illumina (NEB) and then purified by QIAquick PCR Purification Kit (QIAGEN). Afterward, the acquisitions from above were subsequently put into the gel prep for size selection. In addition, the quality of the library was determined using Bioanalyzer 4200 (Agilent). Then, the mRNA-Seq libraries were run in HiSeq X10 system (Illumina) on a paired-end 150-bp run, and the miRNA-Seq libraries were sequenced on an Illumina HiSeq 2500 next-generation sequencer using single-read 50-bp sequencing. Among the compared groups, those with more than 1.2-fold differences of the geometrical mean expression and a statistically significant P value of less than 0.05 based on analysis of DESeq2 were selected as the differentially expressed genes. To determine statistically enriched GO categories, we conducted the following GO analysis of those selected genes using the clusterProfiler R package with P values of less than 0.05. Based on Kyoto Encyclopedia of Genes and Genomes Database (http://www.genome.jp/kegg/), pathway analysis was used to determine the significant pathway of those differential genes. mRNA-Seq and miRNA-Seq data are deposited in the Gene Expression Omnibus repository (GSE142517).

qRT-PCR. All the RNA was extracted using Trizol reagent (Invitrogen). Reverse transcription of miRNA to cDNA was implemented with miRNA-specific primers (Applied Biosystems) preceding the qPCR (TaqMan; Applied Biosystems). Noncoding small RNA control U6 (Applied Biosystems) was adopted and served as an endogenous reference gene, with changes in expression calculated by the change in threshold (ΔΔCT) method. For mRNA, cDNA was prepared by reverse transcription (PrimeScript RT Reagent Kit; Takara). The same amounts of cDNA were adopted for the following qPCR conducted with the SYBR PrimeScript Reverse-Transcription PCR kit (Takara). Amplification was implemented through using an ABI PRISM 7900 Real-Time PCR System (Applied Biosystems). The amplification efficiency of these genes was the same as that for β-ACTIN or RPL13a, as indicated by the standard curves for amplification, allowing us to adopt the subsequent formula, fold difference = 2^{–(ΔCtA −ΔCtB)}, where Ct is the cycle threshold. Primer sequences are presented in Supplemental Table 3.

miRNA target prediction. Targetscan version 7.1 (http://www.targetscan.org/mmu_71/) was adopted to predict the miR-148a targets. The detailed descriptions of the complete computational protocol are available at the website and in reference (47).

Luciferase assay. For 3′UTR reporter assays, the 3′UTR of the MAFB mRNA containing WT or mutant miR-148a binding sites were subcloned into the psiCHECK2 vector (Promega). Then, 1 × 10⁵ HEK 293T cells per well were transfected with a mixture of 100 ng 3′UTR luciferase reporter vector and agomir-miR-148a or agomir control using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions. Twenty-four hours after transfection, cells were lysed, and luciferase activity was measured on a luminometer (Promega) using the Dual-Luciferase Reporter Assay System (Promega). The ratio of renilla luciferase activity to firefly luciferase activity was calculated for each well.

For promoter reporter assay, 200 bp upstream of the TSS of miR-148a with or without a mutant PU.1 binding sequence were subcloned into the pGL3-basic vector (Promega). Then 1 × 10⁴ HeLa cells per well were transfected with a mixture of 100 ng pGL3 luciferase vector and 8 ng pRL-TK Renilla vector with or without 100 ng PU.1 expression plasmid or empty control plasmid. Twenty-four hours after transfection, cells were lysed, and luciferase activity was measured on a luminometer (Promega) using the Dual-Luciferase Reporter Assay System. The ratio of firefly luciferase to renilla luciferase was calculated for each well.

ChIP assays. ChIP assays were conducted with a SimpleChIP Plus Enzymatic Chromatin IP (Magnetic Beads) Kit (Cell Signaling Technology). According to the protocol, monocytes were fixed in 1% formaldehyde and then lysed for 10 minutes on ice. Chromatin was sheared by enzyme digestion and sonication. Then, the cell lysate was immunoprecipitated overnight at 4°C with a monoclonal antibody anti-PU.1 (Cell Signaling Technology, 2258). Subsequently, protein G magnetic beads were added, and the obtained mixture was rotated for 2 hours at 4°C. After washing and elution, cross-links were reversed by heating at 65°C for 2 hours. The eluted DNA was purified, and the samples were analyzed by qPCR with input DNA (total chromatin) as an endogenous control. The promoter in CD11b and downstream of miR-148a TSS region were set as positive or negative control, respectively. ChIP-qPCR primer sequences are listed in Supplemental Table 3.

Ago2 immunoprecipitation. Ago2 immunoprecipitation was carried out according to the protocol provided in the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, 17-700). For the RNA-binding protein immunoprecipitation (RIP) lysate preparation, 2 × 10⁷ human primary monocytes were collected by centrifugation at 211 g for 5 minutes and then lysed with RIP lysis buffer. For the magnetic bead and antibody preparation, purified antibodies (5 μg) were incubated with beads for 30 minutes at room temperature using anti-Ago2 antibody (Millipore, 03-110) as the antibody of interest and normal mouse IgG as negative controls, respectively. Supernatant from the RIP lysate was obtained by centrifugation at 12,210 g for 10 minutes, 100 μL of which was incubated overnight at 4°C with the bead–antibody complex, and 10 μL served as 10% input for generating a standard curve or for qRT-PCR comparison. Immunoprecipitated RNAs were purified and analyzed by qRT-PCR. RNA immunoprecipitation qPCR primer sequences are listed in Supplemental Table 3.

Immunoblots. Cell lysates were prepared by lysing 2 × 10⁶ monocytes or moDCs in Radioimmune Precipitation Assay Lysis and Extraction Buffer (Thermo Fisher Scientific) and then separated with 10% SDS-PAGE. The separated proteins were transferred to an Immobilon-P PVDF membrane (Millipore), which was subsequently blocked with SuperBlock T20 PBS blocking buffer (Thermo Fisher Scientific). Finally, the membrane was incubated with rabbit antibodies against MAFB, 1:1000 (Abcam, ab66506), PU.1, 1:1000 (Cell Signaling Technology, 2258), and β-ACTIN, 1:2000 (Cell Signaling Technology, 4967).

Cytokine detection. Mouse serum was collected at specific time points, and the protein levels of IL-23, IL-17, IL-1β, IL-6, and TNF-α were determined with BD Cytometric Bead Array Mouse/Rat Soluble Protein Master Buffer Kit and Flex Sets, according to the instruction manual provided by the manufacture. Briefly, the phycoerythrin-conjugated detection antibodies are mixed with the specific capture beads and then incubated with recombinant protein standards or samples to generate sandwich complexes. Subsequently, the sandwich complexes of each sample ran on a flow cytometer. Standard curves for all analyses were generated through adopting recombinant protein standards to run one standard mixture. Finally, the sample results were analyzed by FCAP Array software (BD Biosciences).

Statistics. Considering the clinical data, nonparametric Mann-Whitney U test was used to draw comparisons between patients and controls, and Spearman’s test was adopted to carry out the correlation analysis. For all other experiments, data were analyzed through using Graph Prism 7.0 software (GraphPad), and P values were determined by 2-tailed unpaired t test or 1-way ANOVA. P values of less than 0.05 were considered significant.

Study approval. Human sample acquisitions were approved by the Review Board of Renji Hospital of SJTUSM and Huashan Hospital of Fudan University. All experiments were conducted in accordance with the declaration of Helsinki Principles. Informed consent was obtained for all procedures. All the animal experiments were performed with the use of protocols approved by the Institutional Animal Care and Use Committee at Renji Hospital of SJTUSM.