RNA-Editing Enzyme ADAR1 p150 Isoform Is Critical for Germinal Center B Cell Response

Base modifications of eukaryotic mRNAs, also called epitranscriptomic modifications, play multifaceted roles in regulating the fate of individual transcripts, including alternative splicing, recoding, mRNA decay, and translation (1). Adenosine-to-inosine (A-to-I) deamination is one of the first epitranscriptomic modifications identified. It can distinguish cellular RNAs from foreign RNAs, thus preventing overwhelming innate immune responses to the endogenous RNAs (2–4). In mammals, A-to-I RNA editing is catalyzed by two main enzymatically active adenosine deaminases acting on the RNA (ADARs), ADAR1 and ADAR2, both containing dsRNA-binding domains (dsRBDs) in the central region and a deaminase domain in the C terminus (5). In addition, ADAR1 has an N-terminal z-DNA–binding domain (6), which is absent in ADAR2. The ADAR1 protein is expressed as two isoforms, the longer p150 isoform and the shorter p110 isoform. The p150 isoform is IFN inducible and can mediate A-to-I editing in both the nucleus and the cytoplasm (7, 8), whereas the p110 isoform is an N-terminal truncated version of the p150 isoform and constitutively expressed and primarily localized in the nucleus (7, 8). In contrast, ADAR2 is predominantly found in the nucleus, especially the nucleolus (8).

Previous studies have demonstrated that ADAR1 is a critical player in preventing the hyperactivation of multiple dsRNA sensors, including melanoma differentiation-associated protein 5 (MDA5), protein kinase R (PKR), and RNase L, which detect the foreign nucleic acids and initiate the production of type I IFNs (2–4, 9). Human ADAR1 mutations cause Aicardi-Goutières syndrome due to a severe spontaneous IFN response (10). Similarly, ADAR1 deletion in mice leads to aberrant IFN production and embryonic lethality (3, 11, 12). Deleting MDA5 or mitochondrial antiviral signaling protein (MAVS) could partially rescue the embryonic lethality phenotype of ADAR1-deficient mice, but the mice still exhibited multiple-organ developmental failures (3, 4). The MDA5 deficiency could also fully rescue the dysregulated IFN-stimulated gene expression and the embryonic lethality phenotype of Adar1^E861A/E861A mice that were defective in dsRNA editing (13). These studies suggest that ADAR1 plays critical roles in both MDA5–MAVS pathway–dependent and –independent manners and that its RNA-editing activity is required to prevent MDA5 sensing of endogenous dsRNA and spontaneous IFN production. Moreover, ADAR1 has been shown to dampen global translational shutdown and cell death by inhibiting PKR activation during the type I IFN response (2). Additionally, ADAR1 can prevent endogenous dsRNA from activating RNase L and inhibit cell death (9).

In addition to its pivotal functions in embryonic development and innate immunity, ADAR1 is also critical for adaptive immunity. Previous studies showed that Adar1^−/−Mavs^−/− and Adar1^Δ7–9Mavs^−/− mice lacked organized lymphoid follicles in both spleens and lymph nodes, and that the frequency of mature B cells in the spleens was dramatically reduced (3, 12). When ADAR1 was ablated specifically in T cell lineage, the mutant mice exhibited abnormal thymic T cell development, including loss of TCR expression, failure of cell transition, impaired negative selection, and autoimmunity (14, 15). In addition, Marcu-Malina et al. (16) reported that CD19-Cre–mediated Adar1 ablation results in a significant decrease in immature B and mature B cells due to excessive apoptosis, suggesting that ADAR1 is important for late-stage B cell development. However, it remains unknown whether ADAR1 plays a critical role in adaptive immune responses such as the T cell–dependent (TD) Ab response.

During the TD Ab response induced by proteinous Ags, Ag-specific B cells and CD4⁺ Th cells that detect the cognate Ag presented on dendritic cells interact at the T cell/B cell interface. Later, the cognate B and T cells move to the center of the B cell follicle and form a specialized microstructure, called the germinal center (GC). Even prior to GC formation, the activated B cells already initiate Ig class-switch recombination to rearrange their Ig H chain genes (17). Within the GC, B cells undergo somatic hypermutation catalyzed by activation-induced cytidine deaminase (AID) (18). GC B cells shuttle between the dark zone (DZ) and light zone (LZ), facilitated by their varying CXCR4 and CXCR5 expression, and they undergo multiple rounds of proliferation, mutation, and selection (19). Through the GC reaction, the activated B cells differentiate into either long-lived Ab-producing plasma cells or memory B cells, providing an individual with lifelong humoral immunity to protect against microbe infection (20).

In this study, we determined the role of ADAR1 in the TD Ab response. We generated Adar1^f/fAicda^Cre/+ mice, in which the floxed Adar1 gene was specifically ablated by AID-Cre in the activated B cells when AID was inducibly expressed upon immunization. We found that ADAR1 is essential for Ag-specific IgG Ab production and the GC B cell response. We also demonstrated that both MDA5-dependent and -independent pathways are involved downstream of ADAR1 during the GC response. Interestingly, we found that both of the ADAR1 p150 and p110 isoforms are upregulated in B cells upon stimulation, although the p150 isoform is more predominantly expressed than the p110 isoform. More importantly, we revealed that the p150 isoform is exclusively responsible for ADAR1’s role in the GC B cell response. Also, the ADAR1 p110 isoform cannot substitute for the p150 isoform, even with the same cellular localization as the p150 isoform.

Adar1^f/f (21), Cd19^Cre/+ (22), Ifih1^−/− (23), and Aicda^Cre/+ (24) mice were obtained from The Jackson Laboratory. p150^KI/+ mice and p110^KI/+ mice were generated by knocking in the CAG-loxP-STOP-loxP-p110/p150 cassette at the Rosa26 locus of C57BL/6 mice by CRISPR/Cas9-mediated genome engineering. Briefly, two homology arms, a CAG promoter, a STOP region flanked by loxP sites, and a mouse Adar1 coding sequence (p150 or p110) were amplified by PCR and assembled into a targeting vector together with recombination sites and selection markers. The inserted fragments were verified by restriction enzyme digestion and sequencing. The guide RNA (5′-GGATTTAGCCACATCCATAGTGG-3′), the targeting vector containing the mouse Adar1 gene (p150 or p110), and Cas9 mRNA were coinjected into fertilized mouse eggs to generate targeted knock-in (KI) offspring. Eif2ak2^−/− and Rnasel^−/− mice were obtained from Cyagen Biosciences. All strains were maintained on a C57BL/6J background. All mice were housed and bred under a specific pathogen-free condition in the Laboratory Animal Research Center of Southern University of Science and Technology. Six- to 14-wk-old male or female mice were used for experiments, and all animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee of Southern University of Science and Technology.

For the TD immune response, mice were injected i.p. with 100 μg of 4-hydroxy-3-nitrophenyl-acetyl (NP)₃₀–chicken γ-globulin (CGG) (Biosearch Technologies) in Imject alum adjuvant (Thermo Scientific).

Cells were harvested from the spleen or bone marrow (BM) and incubated with RBC lysis buffer (BioLegend) for 3 min on ice. After washing, the cells were stained with corresponding Abs conjugated with various fluorescent dyes or biotin, including anti-B220 (552094, BD Biosciences), anti–BP-1 (553735, BD Biosciences), anti-CD24 (562360, BD Biosciences), anti-CD38 (562770, BD Biosciences), anti-IgG1 (553443, BD Biosciences), anti-CD138 (553713, BD Biosciences), anti-CD21 (562797, BD Biosciences), anti-CD19 (557655, BD Biosciences), anti-Fas (17-0951, eBioscience), anti-IgM (25-5790-82, eBioscience), anti-IgD (13-5993, eBioscience), anti-CD38 (17-0381, eBioscience), anti-CXCR4 (12-9991, eBioscience), anti-CD83 (11-0831, eBioscience), anti-Gr1 (13-5931, eBioscience), anti-streptavidin (45-4317, eBioscience), anti-CD93 (13-5892, eBioscience), anti-CD43 (11-0431, eBioscience), anti-IgM (1140-08, SouthernBiotech), anti-CD3 (100244, BioLegend), and anti-CD23 (101620, BioLegend). The cells were then resuspended in FACS buffer (5% FBS in PBS) and analyzed using CytoFLEX (Beckman Coulter). PE-conjugated (4-hydroxy-5-indo-3-nitrophenyl)acetyl (NIP)-BSA (Biosearch Technologies) was prepared using an R-PE labeling kit-SH (Dojindo) as described previously (25). All Abs were used in 1:100 dilution in FACS buffer or as specified. Dead cells were excluded electronically using DAPI staining.

For ELISPOT, 96-well cellulose ester membrane plates (Millipore) were coated with 0.5 μg/well NP₂₀-BSA in PBS and then blocked with 2% BSA in PBS. Splenocytes and BM cells were added to the well and incubated for 2 h at 37°C in 5% CO₂. After washing, the plates were further incubated with biotin-conjugated anti-IgG1 (17-4015-82, SouthernBiotech) and anti-IgM (1140-08, SouthernBiotech) Abs at 1:1000 dilution followed by streptavidin-conjugated alkaline phosphatase (7100-04, SouthernBiotech) at 1:1000 dilution. Spots were then visualized using the substrate (Mabtech, 3650-10).

For ELISA, 384-well flat-bottom plates were coated with 0.1 μg/well NP₂₀-BSA in PBS and blocked with 2% BSA in PBS. Diluted serum samples collected at indicated time points were added to the well and incubated for 2 h at room temperature. The plates were further incubated with biotin-conjugated anti-IgG1, anti-IgG2b (1186-08, SouthernBiotech), anti-IgG3 (1191-08, SouthernBiotech), and anti-IgM at 1:1000 dilution followed by streptavidin-conjugated HRP (7100-05, SouthernBiotech) at 1:1000 dilution. The plates were visualized using tetramethylbenzidine substrate (421101, BioLegend). OD at 450 nm was measured with a BioTek ELx800 reader (BioTek).

Total RNA was extracted from mouse splenic naive B cells using RNAiso Plus (Takara) and reversely transcribed into cDNAs using a PrimeScript first-strand cDNA synthesis kit (Takara). The cDNAs of mouse ADAR1 p150 and ADAR1 p110 were amplified by PCR and cloned into the MIGR1 retroviral vector. The cDNA of an ADAR1 p150 fragment (aa 120–160) containing the nuclear export sequence (NES) was amplified by PCR and cloned downstream of the start codon of p110 cDNA to generate NES-ADAR1 p110 expression vector. The cDNA of the ADAR1 p150 E861A mutant (13) and ADAR1 p150 EAA mutant (26) were generated by overlap extension PCR and cloned into the MIGR1 retroviral vector. The sequences of the inserts were confirmed by Sanger sequencing. Retroviral package and transduction were performed as previously described (27). Briefly, virus particles were packaged using HEK293FT cells, and the viral supernatants were collected and filtered by a 0.45-μm membrane 2 and 3 d after transfection. Mice were injected i.p. with 3 mg of 5-fluorouracil (Sigma-Aldrich) in 300 μl of PBS. BM cells were harvested 3 d later and cultured in DMEM (HyClone) supplemented with 20% FBS (Life Technologies), 1% penicillin/streptomycin (HyClone), 1% nonessential amino acids (Life Technologies), IL-3 (20 ng/ml, R&D Systems), IL-6 (50 ng/ml, R&D Systems), and stem cell factor (50 ng/ml, R&D Systems) for 24 h. Cells were then spin-infected (2500 rpm for 90 min at 37°C) twice with packaged retroviruses during 2 d. After transduction, cells were 1:2 mixed with μMT BM cells and i.v. injected into lethally irradiated (900 rad) wild-type (WT) recipient mice. The mice were analyzed 6 wk later.

Mouse splenic B cells were purified using anti-CD43 microbeads (Miltenyi Biotec) and cultured at the concentration of 2 × 10⁶ cells/ml in DMEM (HyClone) supplemented with 10% FBS (Life Technologies), 1% penicillin/streptomycin (HyClone), and 1% nonessential amino acids (Life Technologies) at 37°C with 5% CO₂. For RNA sequencing (RNA-seq), B cells were treated with anti-IgM (10 μg/ml, Jackson ImmunoResearch, 115-006-075) and anti-CD40 (1 μg/ml, BD Biosciences, 553721) for 36 h.

Total RNA was extracted from mouse splenic naive B cells, preactivated B cells, and GC B cells using RNAiso Plus (Takara) and reversely transcribed to produce cDNA using HiScript II Q RT SuperMix for quantitative PCR (Vazyme). Quantitative RT-PCR was performed on a qTOWER3 real-time PCR thermal cycler (Analytik Jena) using ChamQ Universal SYBR quantitative PCR master mix (Vazyme). The primers used are as follows: β-actin forward, 5′-CGTGAAAAGATGACCCAGATCA-3′, reverse, 5′-CACAGCCTGGATGGCTACGT-3′; Adar1 forward, 5′-GTTGACGCACTTCCTACAGC-3′, reverse, 5′-AGCAAATAGCACGGGTCAGA-3′; Adar1-p110 forward, 5′-GAAGACTACGCGTTGGGACT-3′, reverse, 5′-GTGTCTGGTGAGGGAACACC-3′; Adar1-p150 forward, 5′-TGTCTCAAGGGTTCAGGGGA-3′, reverse, 5′-TCCTAGGGTAAGACTCCGGC-3′; Ifng forward, 5′-CAATCAGGCCATCAGCAACA-3′, reverse, 5′-GAGCTCATTGAATGCTTGGCG-3′; Cxcl10 forward, 5′-GCCCACGTGTTGAGATCATTG-3′, reverse, 5′-CTCTGCTGTCCATCCATCGC-3′; Isg15 forward, 5′-AAGCAGCCAGAAGCAGACTC-3′, reverse, 5′-CACGGACACCAGGAAATCGT-3′; Ifit1 forward, 5′-CCAGAGAACAGCTACCACCTT-3′, reverse, 5′-TGTGCATCCCCAATGGGTTC-3′; Ifit2 forward, 5′-GTAGGGGTTACATCCGGCAC-3′, reverse, 5′-TCTGTGCAGCACCTCTAAGTC-3′; Rsad2 forward, 5′-CCTGTGCGCTGGAAGGTTT-3′, reverse, 5′-TTCAGGCACCAAACAGGACA-3′; Mx1 forward, 5′-CAGTCATCAGAGTGCAAGCG-3′, reverse, 5′-TCTCCCTCTGATACGGTTTCCT-3′; Oas2 forward, 5′-TGACATGGTGGGAGTGTTCA-3′, reverse, 5′-AGCGTCTTCCAGAGCTGAAT-3′.

Total RNA was extracted using an RNeasy mini kit (Qiagen), and poly(A) mRNA was isolated with the poly(A) mRNA magnetic isolation module (NEB). Libraries were prepared using an NEBNext Ultra RNA library prep kit for Illumina (NEB) and sequenced on an Illumina HiSeq instrument (Illumina). Sequence reads were filtrated using Cutadapt (v1.9.1) and then mapped to the reference genome sequences (GRCm38.98) using HISAT2 (v2.0.1). The differential gene expression analysis was performed using the DESeq2 Bioconductor package, and the differentially expressed genes were screened by an adjusted p <0.05 and a fold change >2. Gene Ontology (GO) enrichment of differentially expressed genes was performed using goseq (v1.34.1).

Splenic B cells were purified using anti-CD43 microbeads (Miltenyi Biotec) and stimulated with LPS (20 μg/ml, Sigma-Aldrich, L2630), anti-IgM (10 μg/ml, Jackson ImmunoResearch, 115-006-075), anti-CD40 (1 μg/ml, BD Biosciences, 553721), or anti-IgM/CD40 for 24 h. Cells were suspended in RIPA buffer (Sigma-Aldrich) containing proteinase inhibitor (Roche) and incubated on ice for 10 min. The lysates were clarified by centrifugation at 8000 × g for 10 min and boiled for 10 min before loading on 10% SDS-PAGE. After electrophoresis, total protein was transferred onto polyvinylidene difluoride membranes followed by blocking with 5% skim milk in TBST for 1 h at room temperature. Membranes were probed with anti-ADAR1 (Ab226188, Abcam), which can bind to the common C terminus region of p150 and p110, or anti–β-actin (66009-1-Ig, Proteintech) Abs at 4°C overnight and further incubated with HRP-conjugated anti-rabbit IgG (TransGen Biotech, HS201-01) at room temperature for 1 h. Blots were developed using chemiluminescent HRP substrate (Millipore) and visualized on a Tanon 5200 imaging system (Tanon).

The cDNAs encoding ADAR1 p150-hemagglutinin (HA) (an HA tag was fused to the C terminus of ADAR1 p150) and NES-ADAR1 p110-FLAG (a FLAG tag was fused to the C terminus of NES-ADAR1 p110) were cloned into retroviral vector MIGR1. HEK293T cells were grown on coverslips and then transfected with ADAR1 p150-HA and NES-ADAR1 p110-FLAG expression vectors. After culturing for 2 d, cells were then fixed with 4% formaldehyde for 30 min at room temperature and blocked with 3% FBS in PBS. Cells were further stained with mouse anti-FLAG Ab (Invitrogen, MA1-91878) and rabbit anti-HA Ab (Cell Signaling Technology, 3724) followed by goat anti-rabbit Ab conjugated with Alexa Fluor 555 (Invitrogen, A21428) and goat anti-mouse Ab conjugated with iFluor 647 (HuaBio, HA1127). After washing, the samples were covered with ProLong Gold antifade reagent with DAPI (Cell Signaling Technology), and immunofluorescence images were taken on a Nikon A1R confocal microscope.

RNA-seq data reported in this study have been deposited in the Gene Expression Omnibus database with the accession number GSE196878 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE196878).

Data are displayed as mean ± SD. Statistical analysis was performed with GraphPad Prism 6 software. An unpaired two-tailed Student t test was used for comparison between two groups, and a one-way ANOVA/Fisher’s least significant difference test was applied for multiple comparisons.

Previous studies demonstrated that the mature B cell population was drastically diminished in Adar1^f/fCd19^Cre/+ mice (16), making these mice not suitable for investigating ADAR1’s role in the Ab response where naive mature B cells are activated by foreign Ags. Therefore, we generated Adar1^f/fAicda^Cre/+ mice, in which the floxed Adar1 alleles are deleted by Cre-recombinase only when AID is inducibly expressed in B cells upon immunization (24). We first determined AID-Cre–mediated deletion of Adar1 gene by quantitative RT-PCR analysis of FACS-sorted activated (CD19⁺CD38⁻Fas⁺) B cells from the Peyer’s patches of Adar1^f/fAicda^Cre/+ and Adar1^f/fAicda^+/+ littermate control mice. We found that the Adar1 mRNA level was decreased by >70% in the activated Adar1^f/fAicda^Cre/+ B cells compared with the control cells (Supplemental Fig. 1A), suggesting that Adar1 is efficiently deleted in the activated mutant B cells.

We next characterized the B cell development in Adar1^f/fAicda^Cre/+ mice without immunization. First, we dissected various B cell subsets in the BM using the Hardy classification and found that the Adar1^f/fAicda^Cre/+ and control mice had equivalent frequencies and numbers of B220⁺CD43⁺CD24⁻BP-1⁻ [Fraction (Fr.) A], B220⁺CD43⁺CD24⁺BP-1⁻ (Fr. B), B220⁺CD43⁺CD24^loBP-1⁺ (Fr. C), B220⁺CD43⁺CD24^hiBP-1⁺ (Fr. C′), B220⁺CD43⁻IgM⁻IgD⁻ (Fr. D), and B220⁺CD43⁻IgM⁺IgD⁻ (Fr. E) B cells (Fig. 1A, 1B). We further examined peripheral B cell development in the spleen of these mice. We found that the frequencies and numbers of the total B220⁺IgM⁺, B220⁺CD93⁻CD21^hiCD23⁻ marginal zone, and B220⁺CD93⁻CD21^loCD23⁺ follicular B cells were largely normal in the Adar1^f/fAicda^Cre/+ mice compared with the control mice (Fig. 1C, 1D). These results suggest that Adar1^f/fAicda^Cre/+ mice have normal B cell development in the BM and peripheral lymphoid organs and can be used as a better model to interrogate ADAR1’s role in the Ab response.

Subsequently, we challenged the Adar1^f/f Aicda^+/+ (control) and Adar1^f/fAicda^Cre/+ mice with a TD Ag, NP, coupled to CGG and measured their sera levels of the NP-specific Abs at various time points postimmunization. Interestingly, we detected substantially equivalent titers of NP-specific IgM in control and Adar1^f/fAicda^Cre/+ mice across the time points examined, except for day 21 postimmunization, when the mutant mice had lower levels of NP-specific IgM compared with the control mice. However, the Adar1^f/fAicda^Cre/+ mice exhibited reduced titers of NP-specific IgG1, IgG2b, and IgG3 Abs at all of the time points postimmunization, except for the IgG2b levels, which were indistinguishable statistically between the mutant and control mice at day 28 postimmunization (Fig. 1E). Thus, these data suggest that ADAR1 is required for the production of Ag-specific IgG Abs during the TD immune response.

We moved on to examine the GC response, which is essential for the TD Ab response. Our flow cytometric analysis revealed that Adar1^f/fAicda^Cre/+ mice had a significant decrease (2- to 3-fold) in the percentage and the absolute number of B220⁺Fas⁺CD38⁻ GC B cells compared with the control (Adar1^f/fAicda^+/+) mice on day 10 postimmunization (Fig. 2A). However, when examining the abundances of DZ and LZ GC B cells based on the expression of CD83 and CXCR4 (28), we found the percentages of DZ and LZ GC B cells to be indistinguishable between the mutant and control mice (Fig. 2B). We also detected the Ag-specific IgG1 B cells on day 10 postimmunization. Corroborating the decrease in total GC B cells, the mutant mice had reduced Ag-specific NIP⁺IgG1⁺ B cells compared with the control mice (Fig. 2C).

Furthermore, the mutant mice also had a diminished Ag-specific CD38^hiIgG1^hi memory B cell population. The percentage and the absolute number of NIP⁺CD38^hiIgG1^hi memory B cells were decreased by >60% and >90%, respectively, in the mutant compared with the control mice at day 10 postimmunization (Fig. 2D). We also measured the frequency of NP-specific IgG1 Ab-secreting cells (ASCs) by an ELISPOT assay. We found that the mutant mice had significantly reduced NP-specific IgG1 ASCs in the spleen and BM compared with the control mice at day 10 postimmunization (Fig. 2E, 2F). Taken together, our results indicate that ADAR1 is required for the GC B cell response and the generation of Ag-specific IgG1 memory B cells and ASCs. We also compared the GC phenotypes of Adar1^f/fAicda^Cre/+ mice using Adar1^+/+Aicda^Cre/+ mice as the control. We consistently detected almost the same defects in Adar1^f/fAicda^Cre/+ mice when Adar1^+/+Aicda^Cre/+ mice were used as the control (Supplemental Fig. 1B–F). Thereafter, we used Adar1^f/fAicda^+/+ instead of Adar1^+/+Aicda^Cre/+ mice as the littermate controls for the following experiments, as it is easier to get enough numbers of Adar1^f/fAicda^Cre/+ and Adar1^f/fAicda^+/+ littermates by crossing Adar1^f/fAicda^Cre/+ with Adar1^f/fAicda^+/+ mice.

We proceeded to investigate the underpinning mechanisms whereby ADAR1 regulates the GC B cell response. To this end, we performed RNA-seq analysis on ADAR1-deficient B cells stimulated in vitro with anti-IgM and anti-CD40 Abs, which mimics the B cell activation in the GC response. We observed that among the 263 genes differentially expressed (fold change ≥ 2, p ≤ 0.05), 244 genes (∼92%) were significantly upregulated whereas only 19 genes (∼8%) were downregulated in the mutant compared with the control B cells after stimulation (Fig. 3A). Notably, GO enrichment analysis revealed that the genes involved in the antiviral response, dsRNA binding, and the type I IFN response were enriched in the ADAR1-deficient B cells (Fig. 3B). These results suggest that the activated ADAR1 mutant B cells have an exacerbated antiviral IFN response due to excessive sensing of the unedited endogenous dsRNAs.

As the MDA5–MAVS pathway is hyperactivated and primarily responsible for the defects in ADAR1 knockout (KO) mice (3, 4, 12), we next asked whether concurrently deleting MDA5 could rescue the GC B cell defects in the Adar1^f/fAicda^Cre/+ mice. Thus, we crossed Adar1^f/fAicda^Cre/+ with Ifih1^−/− mice to generate Adar1^f/fAicda^Cre/+Ifih1^−/− (ADAR1/MDA5 double KO [DKO]) mice. Subsequently, we immunized the mice with NP-CGG and examined the GC response on day 10 postimmunization. We found that the concurrent MDA5 deletion could lead to an ∼50% and 80% increase in the percentage and the number of GC B cells, respectively, in the spleen of DKO mice compared with ADAR1-deficient mice (Fig. 3C). These results suggest that the hyperactivation of the MDA5 pathway triggered by the excess endogenous dsRNAs is partially responsible for GC B cell defects in the ADAR1-deficient mice. However, the percentages of Ag-specific IgG1 B cells and ASCs in the DKO mice were not increased as examined by flow cytometry and an ELISPOT assay (Fig. 3D, 3E). Interestingly, we also noticed that Ifih1^−/− mice had a reduced frequency of GC B cells compared with the control mice, albeit the difference in the total GC B cell numbers between the two strains was not statistically significant (Fig. 3C). A detailed analysis of GC B cells revealed that the Ifih1^−/− mice had a higher DZ/LZ B cell ratio compared with the WT mice (Supplemental Fig. 2). These subtle differences imply that MDA5 might play a positive role in regulating the GC response, possibly in an ADAR1-independent manner. Nevertheless, our results suggest that ADAR1 regulates the GC response by preventing the hyperactivation of the MDA5–MAVS pathway.

PKR and RNase L are the other two pathways downstream of ADAR1 (2, 9). The previous study has shown that concurrent deletion of PKR failed to rescue cell apoptosis and embryonic lethality in ADAR1 KO mice (21). However, another study recently demonstrated that PKR is involved in the disease progression caused by ADAR1 dysfunction and loss of PKR could protect ADAR1 mutant mice from mortality and weight loss in vivo (29). Therefore, to assess whether the impaired GC B cell response in the ADAR1-deficient mice was due to the hyperactivation of PKR, we first crossed Adar1^f/fAicda^Cre/+ with Eif2ak2^−/− mice to generate Adar1^f/fAicda^Cre/+Eif2ak2^−/− (ADAR1/PKR DKO) mice. When examining the GC response on day 10 postimmunization, we noticed that both ADAR1 KO and ADAR1/PKR DKO mice had comparable percentages and numbers of B220⁺Fas⁺CD38⁻ GC B cells in the spleen (Fig. 4A). Furthermore, they also had indistinguishable frequencies of splenic NP-specific IgG1 ASCs as assessed by an ELISPOT assay (Fig. 4B).

RNase L is another key effector that mediates cell death downstream of ADAR1 (9, 30). RNase L is activated by binding to 2′-5′-oligoadenylate, a product of the dsRNA-activated 2′-5′-oligoadenylate synthetase, and then cleaves nonspecific cellular RNAs, leading cells to apoptosis (31). Interestingly, GO enrichment analysis showed that the genes involved in 2′-5′-oligoadenylate synthetase activity were enriched in ADAR1-deficient B cells (Fig. 3B). Therefore, we further examined the contribution of the RNase L pathway to the GC B cell response in ADAR1 KO mice. We bred the Adar1^f/fAicda^Cre/+ mice with Rnasel^−/− mice to generate Adar1^f/fAicda^Cre/+ Rnasel^−/− (ADAR1/RNase L DKO) mice. After challenging the mice with TD Ag NP-CGG, we examined the GC B cell response in the spleen on day 10 postimmunization. We found that the ADAR1-deficient and DKO mice had a comparable frequency and number of GC B cells in the spleen upon immunization (Fig. 4C). Furthermore, the ADAR1-deficient and DKO mice also had an indistinguishable number of splenic NP-specific IgG1 ASCs (Fig. 4D), suggesting that the RNase L pathway is not involved in the GC B cell response downstream of ADAR1. Collectively, our results indicate that ADAR1 regulates the GC response independently of the PKR and RNase L pathways.

Previous studies have demonstrated that ADAR1 p150 was predominantly expressed in the spleen and thymus, whereas ADAR1 p110 was mainly found in the brain (14). It was also reported that ADAR1 p110 and p150 regulate multiple-organ development in an isoform-specific manner (3). In this study, we attempted to determine the role of the two ADAR1 isoforms in the GC response. First, we examined the expression of the two Adar1 isoforms in different B cell subsets and mature B cells upon various stimulations. We noticed that p150 was more predominantly expressed than p110 in various B cell subsets by analyzing the data from the Immunological Genome Project RNA-seq datasets (GSE109125) (Fig. 5A). The definition of B cell subsets has been described previously (32). Upon treatment with LPS, anti-IgM, and anti-CD40 Abs, the mRNA level of p110 was substantially elevated whereas p150 mRNA was not significantly increased after the same stimulations (Fig. 5B). Interestingly, protein levels of both the p150 and p110 ADAR1 isoforms were considerably upregulated in B cells upon various stimulations, and the abundance of p150 was much higher than p110 (Fig. 5C, 5D), indicating that the expression of p150 protein might be regulated posttranscriptionally or posttranslationally. As many studies have shown that p150 transcription is induced upon IFN stimulation (7, 33), we examined whether anti-CD40 and anti-IgM plus anti-CD40 Ab treatment could upregulate IFN expression and therefore cause the induction of p150. To this end, we examined the expression of IFN and IFN-stimulated genes in B cells treated with anti-CD40 and anti-IgM plus anti-CD40 Abs. We found that only the expression of IFN-γ but not IFN-α or IFN-β was detectable. The IFN-γ expression was unchanged upon anti-IgM plus anti-CD40 Ab stimulation or even decreased upon anti-CD40 Ab stimulation (Supplemental Fig. 3). Similarly, the expressions of IFN-related genes such as Ifit2, Rsad2, Mx1, and Oas2 were also decreased upon anti-IgM plus anti-CD40 Ab stimulation, although the expressions of Isg15, Cxcl10, and Ifit1 were increased upon anti-CD40 but not anti-IgM plus anti-CD40 Ab treatment (Supplemental Fig. 3). These results indicate that the induction of Adar1 in B cells upon various stimuli is independent of IFN and IFN-related genes.

Next, we asked whether ADAR1 p150 and p110 isoforms equally or differentially contribute to the GC B cell response. To this end, we employed a gain-of-function gene-targeting approach to generate Adar1 p150 or p110 KI (p150^KI/+ or p110^KI/+) mice by inserting a cDNA fragment encoding p150 or p110, preceded by the synthetic CAG promoter and a loxP-flanked Neo-STOP cassette, into the ubiquitously expressed Rosa26 locus (Supplemental Fig. 4A). Next, we crossed the p150^KI/+ or p110^KI/+ mice with Aicda^Cre/+ mice to generate p150^KI/+Aicda^Cre/+ or p110^KI/+Aicda^Cre/+ mice, in which the knocked-in p150 or p110 transgene is overexpressed in the activated B cells where AID-Cre recombinase is inducibly expressed to remove the STOP region. Due to the limited number of AID-expressing GC B cells, we purified splenic mature B cells from Adar1^f/fCd19^Cre/+p110^KI/+ (p110 KI) and Adar1^f/fCd19^Cre/+p150^KI/+ (p150 KI) mice to determine the Cre-mediated expression of p150 and p110. Our quantitative RT-PCR analysis showed that the mRNA levels of p150 and p110 in B cells from the KI mice were ∼4- and 2.5-fold of those in the control B cells (Supplemental Fig. 4B). Furthermore, immunoblotting detected significantly enhanced p150 and p110 proteins in the B cells from the KI mice compared with the control B cells (Supplemental Fig. 4C).

Subsequently, we bred the p150^KI/+Aicda^Cre/+ or p110^KI/+Aicda^Cre/+ mice with Adar1^f/f mice to obtain Adar1^f/fp150^KI/+Aicda^Cre/+ (Adar1^cKOp150^cKI) or Adar1^f/fp110^KI/+Aicda^Cre/+ (Adar1^cKOp110^cKI) mice. We further challenged the mice with NP-CGG and examined whether the defective GC B cell response in the Adar1^f/fAicda^Cre/+ mice could be rescued by the p150 or p110 transgene. Our flow cytometry demonstrated that the Adar1^cKOp150^cKI and control mice had comparable GC B cell populations in their spleens on day 10 postimmunization (Fig. 6A). Enumeration of the B220⁺Fas⁺CD38⁻ GC B cells revealed that Adar1^cKOp150^cKI mice had normal, if not slightly increased, percentages and absolute numbers of GC B cells compared with the control mice (Fig. 6B). Consistently, the decreased splenic NP-specific IgG1 B cells in the ADAR1-deficient mice were also successfully restored in the Adar1^cKOp150^cKI mice (Fig. 6C, 6D). In contrast, the Adar1^cKOp110^cKI mice manifested similarly compromised GC- and NP-specific IgG1 B cell populations as those of the ADAR1-deficient mice (Fig. 6A–D). Also, the ELISPOT assay showed that the defective NP-specific IgG1 ASC production levels in the spleen and BM of ADAR1-deficient mice were only rescued by the p150 but not p110 transgene (Fig. 6E, 6F). Taken together, these results suggest that the p150 isoform exclusively accounts for the role of ADAR1 in the GC B cell response.

Although ADAR1 p150 and p110 share the same deaminase domain and three dsRBDs, they exhibit distinct cellular localizations. The p150 possesses an NES domain (aa 122–153), which is absent in the p110 isoform (34). Consequently, the ADAR1 p150 can translocate from the nucleus to the cytoplasm, whereas the p110 is mainly present in the nucleus. Thus, we asked whether the unsuccessful rescue of the GC response in the Adar1^cKOp110^cKI mice was due to p110’s inability to translocate from the nucleus to the cytoplasm where the enzyme can access its substrates. Therefore, we attempted to reintroduce a nucleus-exportable p110 into the Adar1^cKO mice. We generated a chimeric ADAR1 p110 (ADAR1 NES-p110) by fusing the full-length p110 with an NES-containing fragment (aa 120–160) of the p150 (Fig. 7A). The confocal microscopic study demonstrated that the NES-p110 was exclusively localized in the cytoplasm and had the same localization as the p150 overexpressed in HEK293T cells (Fig. 7B). Next, we performed mixed BM chimera reconstitution experiments by transferring 33% of the BM cells of Adar1^cKO mice transduced retrovirally with the WT p110 (KO^+p110) or the NES-p110 (KO^+NES-p110), together with 66% of the BM cells of μMT mice into the sublethally irradiated WT mice. We also made three other chimeras as controls, using 66% of the BM cells of μMT together with 33% of either the control BM cells transduced with the empty vector (control^+empty) or Adar1^cKO BM cells transduced with the empty vector (KO^+empty), or Adar1^cKO BM cells transduced with the WT p150 (KO^+p150).

We immunized the chimeric mice with NP-CGG 6 wk after the BM reconstitution and evaluated the GC B cell response by flow cytometry on day 10 postimmunization. Consistent with the defective GC response in the ADAR1 cKO mice, the KO^+empty chimera displayed significantly reduced GC B cells compared with the control^+empty chimera (Fig. 7C–E). Corroborating the essential role of the p150 in the GC B cell response, the KO^+p150 chimera had completely rescued GC B cells in the spleen. However, neither the KO^+p110 nor the KO^+NES-p110 had a restored GC B cell population as the KO^+p150 chimera, demonstrating that the nucleus-exportable NES-p110 fails to rescue the severely compromised GC B cells in the absence of ADAR1 (Fig. 7C–E). Thus, our results suggest that ADAR1 p110 exerts no function during the GC B cell response, even with the same cellular localization as the p150 isoform.

We further explored the mechanism whereby ADAR1 p150 regulates the GC response by examining the contribution of ADAR1 p150’s RNA-editing activity or the dsRNA-binding activity to its function. We generated a mixed BM chimera by reconstituting ADAR1-deficient B cells with WT p150, an RNA-editing deficient mutant p150^E861A (the glutamic acid at the position 861 was changed to alanine) (13), or an RNA-binding defective mutant p150^EAA (the KKXXK motif was mutated to EAXXA in all three dsRBDs of the p150) (26). Similar to the WT p150, the p150^E861A could completely rescue the GC defects of the ADAR1-deficient B cells in the chimeric mice (Fig. 7F–H), indicating that the RNA-editing activity is dispensable for p150’s role in the GC response. In contrast, the p150^EAA mutant failed to rectify the defects, implying that the dsRBDs are required for p150’s function in the GC response. Thus, our results suggest that the conventional RNA-binding activity but not the RNA-editing activity is essential for ADAR1’s function in regulating the GC B cell response.

Previous studies have reported that ADAR1 is critical for embryonic development and innate immunity by preventing endogenous dsRNAs from triggering nucleic acid sensors such as MDA5, PKR, and RNase L (2–4, 9). In this study, we demonstrate that ADAR1 also plays an essential role in generating TD Ag-specific IgG Abs by positively regulating the GC B cell response. The activated ADAR1-deficient B cells display augmented expression of genes involved in type I IFN antiviral response, indicating an exacerbated sensing of unedited cellular RNA in the activated mutant B cells. In line with these results, concomitant ablation of MDA5 could partially restore the decreased percentage and number of GC but not the Ag-specific IgG1 B cell response in ADAR1-deficient mice. Interestingly, we found that MDA5-deficient mice had a mildly decreased percentage of GC B cells and a higher DZ/LZ GC B cell ratio compared with the control mice, indicating that MDA5 might also be involved in regulating GC B cells. Furthermore, the frequency and the total number of GC B cells were largely comparable between the Ifih1^−/− and the DKO mice. It is likely that ADAR1 might regulate other pathways in addition to MDA5, and at the same time, MDA5 could also have other upstream regulators in addition to ADAR1. In the DKO mice, although the deletion of MDA5 rectified the hyperactivation triggered by the excess endogenous dsRNAs, the positive signals mediated by MDA5 and independent of ADAR1 could also be lost, thus accounting for the partially rescued GC B cell phenotype.

One of the possible MDA5-independent mechanisms downstream of ADAR1 for regulating the GC response could be the regulation of miRNAs. A previous study showed that ADARs have editing-independent effects on the miRNA/small interfering RNA pathways (35). A more recent study demonstrated that ADAR1 forms a heterodimer complex with Dicer through direct protein–protein interaction, facilitating pre-miRNA cleavage by Dicer and miRNA loading onto the RNA-induced silencing complexes (36). In the ADAR1 mutant mouse embryo, the expression of miRNAs is globally shut down. Previously, we showed that global deletion of miRNAs by ablating Dicer in the activated B cells abolishes GC B cell generation (37). Thus, it will be interesting to investigate whether the expression of miRNAs is affected in the ADAR1-deficient GC B cells.

Another interesting finding of the current study is that the p150 isoform of ADAR1 exclusively accounts for ADAR1’s role in the GC B cell response, and the p110 isoform could not replace the role of p150 role in regulating the GC response. We also found the p150 to be more predominantly expressed in resting and activated B cells than the p110, similar to T cells that also primarily express the p150 isoform (14). In contrast, previous studies showed that the p110 isoform is more abundantly expressed than the p150 in tumors such as lung cancer (38). These results suggest that the p150 and p110 ADAR1 isoforms are differentially expressed and might exert distinct functions in different contexts. Interestingly, we also reveal that the protein but not the mRNA level of p150 is significantly elevated in B cells upon various stimulation, indicating that the p150 isoform could be regulated posttranscriptionally through some unknown mechanisms. Indeed, previous studies have shown that the ADAR1 transcript could be directly targeted by miRNAs (39) and ADAR1 protein could be marked for degradation by ubiquitination in tumor cells (40). Moreover, a recent study showed that the translational efficiency of N ⁶-methyladenosine–modified ADAR1 p150 mRNA was significantly enhanced by YTH N⁶-methyladenosine RNA-binding protein 1 (41). However, the detailed mechanisms whereby ADAR1 p150 is regulated posttranscriptionally or posttranslationally in B cells remain unclear and need to be investigated in the future.

The p110 isoform of ADAR1 is the truncated form of p150, bearing the same central dsRBDs and the C-terminal deaminase domain as p150 but lacking the NES and Zα domain at the N terminus. The absence of the NES in the p110 isoform confines it to be localized only in the nucleus. It was plausible that the inability of p110 to access the cytoplasmic RNA targets was the reason for the unsuccessful rescue of the defective GC B cell phenotypes in the ADAR1-deficient mice by the p110 KI. However, our mixed BM chimera experiment using p110-NES showed that this is not the case. The overexpression of a nucleus-exportable p110 cannot substitute for the role of p150 in regulating the GC response. A recent study employing RNA immunoprecipitation sequencing showed that the p150 and p110 ADAR isoforms are associated with different sets of genes, suggesting that they bind to distinct RNA targets (42). Besides the NES, the p150 has an extra Zα domain at its N terminus. Thus, it is tantalizing to hypothesize that the Zα domain might mediate, either by itself or in cooperation with other domains such as the dsRBDs, the binding of its specific targets critical for ADAR1’s function in regulating the GC response. Indeed, we confirm that the dsRNA-binding activity of p150 is essential for its role in the GC B cell response, as the p150^EAA defective in dsRNA binding cannot rescue the defects of ADAR1-deficient GC B cells.

In summary, the current study unveils a critical role of the ADAR1 p150 isoform in the GC B cell response. Our findings shed light on the understanding of the mechanisms whereby GC and Ab responses are regulated, which is helpful for the future development of vaccines and anti-autoimmunity therapy.

The authors acknowledge the assistance of the SUSTech Laboratory Animal Center for the care of mice and the SUSTech Core Research Facilities for training and access to the FACS sorter.

The authors have no financial conflicts of interest.