ADAR1 Is Required for Dendritic Cell Subset Homeostasis and Alveolar Macrophage Function

Editing of RNA by adenosine deaminases acting on dsRNA (ADAR) represents the primary mechanism for adenosine to inosine conversion (A-to-I) and regulates transcription levels of precursor mRNA (1). This mechanism is highly conserved in mammals, with emerging evidence revealing that A-to-I RNA editing is not a rare occurrence but represents a widespread phenomenon modulating a large fraction of the human genome (2–5). To date, three conserved human ADAR family members have been identified (ADAR1, ADAR2, and ADAR3). Whereas ADAR1 is ubiquitously expressed, ADAR2 and ADAR3 are preferentially or exclusively expressed in the brain, respectively (1). Recently, the first global study on ADAR1–RNA interaction using cross-linking immunoprecipitation revealed ADAR1 binding sites in >10,000 protein-coding human genes (6), including a large number of non-Alu ADAR1 binding sites, suggesting additional functional roles of A-to-I editing, such as regulation of 3′ UTRs and microRNA processing.

ADAR1 deficiency is embryonically lethal in mice due to global upregulation of type I IFN–inducible transcripts and rapid apoptosis (7, 8). Importantly, deleting either MDA5 or MAVS rescues the embryonic lethal phenotype of ADAR1-deficient mice (9–11). Moreover, mutations in the human ADAR1 gene or aberrant ADAR1 activities have been associated with severe diseases, including Aicardi–Goutières Syndrome, dyschromatosis symmetrica hereditaria, and amyotrophic lateral sclerosis (12–15). An increasing number of studies have shown that ADAR1 mutations and alterations in RNA editing rates play key roles as drivers of malignancy in various cancers, including leukemia, glioblastoma multiforme, and metastatic melanoma (16–18). The discovery of widespread RNA editing in human protein-coding genes indicates an important role for ADAR1 in cellular development and suggests a novel therapeutic target for various disorders (4). However, the role of ADAR1 in the development and function of dendritic cells (DC), representing critical APCs, remains unknown.

DC differentiate from CD34⁺ bone marrow (BM) progenitor cells and represent a heterogeneous group of professional APCs that play key roles in interactions of innate and adaptive immunity (19–21). Major DC subsets include plasmacytoid DC (pDC), Batf3-dependent CD103⁺ DC, and IRF4-dependent CD11b⁺ DC (22). pDC are specialized in the recognition of virus-derived particles and represent unsurpassed producers of TLR7- and -9-induced IFN-α (23). CD103⁺ DC share functions and ontogeny with CD8⁺ lymphoid DC and exhibit superior capacities to cross-present Ags via MHC class I (MHC-I) molecules to CD8⁺ T cells (22, 24). Independently, CD103⁺/CD8⁺ DC subsets demonstrate superior naive CD8⁺ T cell priming capabilities due to their capacity to produce large amounts of IL-12 and IL-15 and express the chemokine receptor XCR-1, which attracts and binds XCL-1 on activated CD8⁺ cells to promote differentiation of cytotoxic T cells (22, 25, 26). In contrast, CD11b⁺ DC are specialized in MHC class II (MHC-II)–restricted Ag presentation and activation of CD4⁺ T cells (27).

As the loss of ADAR1 is embryonically lethal, we addressed the in vivo roles of ADAR1 in DC development and function using the CD11c-Cre–mediated conditional ablation of ADAR1. Our results revealed that ADAR1 deficiency had systemic effects on DC development, particularly on lung DC as well as alveolar macrophages (AM). More specific, a dramatic loss of pulmonary CD103⁺ DC in comparison with CD11b⁺ DC and pDC subsets was observed. Moreover, systematic investigations of migratory and lymphoid DC subsets indicated preferential depletion of CD103⁺/CD8⁺ DC and a developmental blockage of the cDC1 progenitor lineage, revealing its dependency on ADAR1. ADAR1 deficiency in CD11c⁺ cells also resulted in markedly reduced numbers of AM, which presented lipids and carbohydrate macromolecule accumulations, indicating disturbed surfactant catabolism resembling pulmonary alveolar proteinosis.

In this study, we establish the essential role of ADAR1 in the development of the two key immune cell subsets AM and CD103⁺/CD8⁺ DC, providing insight into the relevance of ADAR1 in DC immunobiology and its cell type–specific effects.

ADAR1^fΔ7-Δ9 floxed mice were provided by M. Higuchi and P. Seeburg (Max Planck Institute for Medical Research, Heidelberg, Germany) (28). CD11c-Cre mice (B6.Cg-Tg[Itgax-cre]1-1Reiz/J) (29) were obtained from The Jackson Laboratory (strain 008068). LysM-Cre mice were provided by I. Forster (Limes Institute, Bonn, Germany) (30). C57BL/6 ADAR1^fΔ7-Δ9/CD11cCre, C57BL/6 ADAR1^wt/wt/CD11c-Cre, or C57BL/6 ADAR1^fΔ7-Δ9/LysMCre mice were generated by breeding of the ADAR floxed mice to the corresponding Cre driver line under specific pathogen-free conditions in the animal facility (BMFZ, Philipps University Marburg, Germany), OVA-specific CD8⁺ (OT-I) (Tg[TcrOva]1100Mjb), and OVA-specific CD4⁺ (OT-II) (Tg[TcraTcrb]425-2Cbn) TCR transgenic mice (Sabine Kranz, Center for Experimental Molecular Medicine, Würzburg, Germany) were maintained under specific pathogen-free conditions at the University of Giessen. Protocols for preparation and analysis of mouse tissues were approved under no. 453_M, no. 621_M, and by the regional animal authority board.

Mice were genotyped by PCR analysis of genomic DNA prepared from tail tips and ear punches. Tail tips and ear punches were lysed in 500 μl of buffer containing 10 mM Tris HCl (Roth), 25 mM EDTA (Roth), 100 mM NaCl (Roth), 0.5% SDS (Roth), and 100 μg/ml proteinase K (Roche) overnight at 56°C. Lysates were centrifuged at 13,000 × g for 10 min, and supernatants were transferred into a separate tube. DNA was precipitated by addition of 500 μl of isopropanol (Roth), removed with a tip, and transferred into a tube containing 300 μl of H₂O. DNA was dissolved for 2 h at room temperature (RT).

ADAR1^fΔ7-Δ9 floxed mice were analyzed for the loxP site located downstream of exon 9 with primer pair 536 (5′-CTGCCACTTCTCCCTGACTC-3′) and 537 (5′-AGTCCTCTCCCTTCCCTGAA-3′). Presence of the loxP site resulted in a 34-bp enlarged PCR product in comparison with unfloxed alleles. The insertion of the Cre transgene was analyzed with primer pair 1035 (5′-ACTTGGCAGCTGTCTCCAAG-3′) and 1036 (5′-GCGAACATCTTCAGGTTCTG-3′) for Cre recombinase under control of the CD11c promoter (31) and primer pair 672 (5′-CCCAAGAAGAAGAGGAAGGTGTCC-3′) and 673 (5′-CCCAGAAATGCCAGATTACG-3′) for Cre recombinase under control of the LysM promoter (30). All primers were purchased from Metabion.

PCRs were carried out in a final volume of 30 μl containing 15 μl of DreamTaq Green PCR Master Mix (2×) (Thermo Fisher Scientific), 1 μl of primer 1 (final concentration 0.33 pmol/μl), 1 μl of primer 2 (final concentration 0.33 pmol/μl), and 100 ng of genomic DNA. The PCR conditions were set as (1) initial denaturation at 95°C for 5 min (2); 35 cycles of amplification including 95°C denaturation for 20 s (flox) or 42 s (LysM-Cre) or 30 s (CD11c-Cre), 55°C (flox, LysM Cre) or 63°C (CD11c-Cre) annealing for 30 s (flox, CD11c-Cre) or 40 s (LysM-Cre) and 72°C extension for 30 s (flox, CD11c-Cre) or for 40 s (LysM-Cre) (3); a final extension at 72°C for 5 min (flox, LysM-Cre) or 10 min (CD11c-Cre). The reactions were performed in a Bio-Rad thermal cycler (Hercules), and PCR products were resolved in a 2% agarose gel. Expected PCR product size were as follows: ADAR1 wild-type (WT) 217 bp, ADAR1 flox 251 bp, LysM-Cre 580 bp, and CD11c-Cre 300 bp.

Mouse lung tissues preparations and bronchoalveolar lavage (BAL) fluid was isolations were conducted as described previously (31). Briefly, mouse lungs were perfused via the right ventricle with HBSS containing 0.35 g/l NaHCO₃ (PAN Biotech) and minced and incubated with lung digestion buffer containing 8.8 U/ml DNase I (Roche) and 0.153 U/ml Collagenase A (Roche) in 10% FCS/RPMI 1640 (PAA Laboratories, PAN Biotech) for 1 h at 37°C. Cells were resuspended in HBSS, mashed through a 70-μm cell strainer (Thermo Fisher Scientific), and washed with HBSS after erythrocyte lysis (155 mM NH₄Cl; 10 mM KHCO₃; 10 mM EDTA; pH 7.4), followed by centrifugation at 400 × g for 5 min at 4°C.

Spleen tissues were injected with lung digestion buffer containing 8.8 U/ml DNase I, 0.153 U/ml Collagenase A (Roche), 10% FCS/RPMI 1640, and were incubated for 20 min at 37°C. Digested tissue was gently mashed through a 70-μm cell strainer, washed with HBSS after erythrocyte lysis, and centrifuged at 400 × g for 5 min at 4°C.

Thymuses together with mesenteric and inguinal lymph nodes were excised, minced, and incubated with lung digestion buffer for 20 min at 37°C, as described previously (32). Cells from respective tissues were resuspended in HBSS, mashed through a 70-μm cell strainer, washed with HBSS, and centrifuged at 400 × g for 5 min at 4°C.

BM was flushed from femora and tibiae using RPMI 1640 supplemented with 10% FCS. BM samples were subsequently filtered through nylon mesh; erythrocytes were lysed with erythrocyte lysis buffer and were then filtered through a 70-μm cell strainer followed by centrifugation (400 × g for 5 min at 4°C).

Skin cells were prepared as previously described (33). Briefly, 4–5 cm² and two mice ear skin pieces were dissected and digested with 3 ml of digestion mixture comprising RPMI 1640, 5 mM HEPES (Life Technologies), collagenase from clostridium histolyticum (9 μg/ml, Sigma-Aldrich), hyaluronidase (0.5 μg/ml, Sigma-Aldrich), and 0.1 mg/ml DNase I for 45 min at 37°C. Digested skins were then minced, washed twice in ice-cold medium containing 10 mM EDTA, and filtered through a 70-μm cell strainer.

Small intestinal tissues were prepared as previously described (34). Briefly, tissue samples were dissected followed by removal of intestinal contents and Peyer patches. Mucosa were removed by washing with HBSS containing 5 mM DTT (Sigma-Aldrich), 2% FCS. Epithelial cells were removed with 5 mM EDTA in HBSS with 2% FCS, followed by washing, mincing, and digestion with intestine digestion buffer (HBSS, 10 mM HEPES, 200 U/ml DNase I, 0.2 U/ml Liberase TM [Roche]). The resulting cells were resuspended, filtered, and washed with HBSS.

Liver tissues were prepared as described previously (35). Briefly, livers were perfused via the inferior vena cava, minced, and incubated in liver digestion buffer (HBSS, 10 mM HEPES, 3 mM CaCl₂, 9.09 U/ml DNase I, 0.2 U/ml Liberase TM) for 30 min at 37°C. Single-cell suspensions were prepared by aspiration and subsequent step-wise filtration through a 100- and 70-μm cell strainer, followed by centrifugation at 400 × g for 5 min at 4°C three times.

To verify the CD11c-dependent loss of the ADAR1^{fΔ7−9/Δ7−9} fragment in CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice, spleen DCs were analyzed for the transcriptional deletion of exon 7 to 9 in ADAR1 mRNA. Spleens were prepared as described above with subsequent CD3/CD19 MagniSort depletion in accordance with the manufacturer’s protocol (Thermo Fisher Scientific). Spleen DC populations were afterward sorted using a BD FACSAria III and defined as CD11b⁺ CD11c⁺ Gr1⁻MHC-II⁺ NK1.1⁻ (CD11b⁺ CD11c⁺ DC), CD11b⁺ CD11c⁻ Gr1⁻ MHC-II⁺ NK1.1⁻ (CD11c⁻ CD11b⁺ DC), SiglecH⁺ CD317⁺ CD11c^+/− MHC-II^+/− CD11b⁻ NK1.1⁻ (pDC), CD8⁺ CD11c⁺ Gr1⁻ MHC-II⁺ NK1.1⁻ (CD8⁺ CD11c⁺ DC; Fig. 1A).

Total RNA was isolated according to the RNeasy Plus Micro Kit (Qiagen) followed by reverse transcription of RNA into cDNA using the QuantiTect Reverse Transcription Kit (Qiagen). cDNA samples were analyzed by quantitative RT-PCR for the presence of the exon 5-7/8 fragment (primer pair 5_7/8: mADAR1_sense_ex5 [5′-TGTGTAGCAGTAGGAGCCCA-3′]; mADAR1_anti_ex7_8 [5′-AGAGAGGAAGCTCTGCGAAAC-3′]) (Metabion) to quantify ADAR1 WT mRNA or for the presence of exon 5/6-11 fragment with deleted exons 7–9 to quantify ADAR1^{fΔ7−9/Δ7−9} mRNA (primer pair 5/6_11: mADAR1_sense_ex5_6 [5′-CAGTTCAGCGGATGACCAGT-3′]; mADAR1_anti_ex11 [5′-ACGGTCACTGCAGGATTTGT -3′]). Quantitative RT-PCRs were performed in a final volume of 20 μl containing 10 μl of PowerUp SYBR Green Master Mix (Thermo Fisher Scientific), primer pair 5_7/8 (final concentration 500 nM) or primer pair 5/6_11 (final concentration 300 nM) and 1 μl of cDNA. PCR conditions were set to 2 min of initial denaturation at 95°C, 40 cycles of amplification with 15 s of denaturation at 95°C, 15 s at 55°C for annealing, and 30 s at 72°C for extension with a final extension at 72°C for 5 min followed by melt curve analysis (60°C to 95°C). Quantitative RT-PCR was performed with the MiniOpticon system (Bio-Rad). Absolute quantification was performed using the standard curve method. Standards were prepared as follows: primer pair 5_7/8 and primer pair 5/6_11 were used to generate PCR products using cDNA from sorted CD11c⁻ CD11b⁺ cells or sorted CD8⁺ CD11c⁺ DCs, respectively. PCR products were purified by agarose gel electrophoresis followed by gel extraction according to the manufacturer’s protocol (GeneJet Gel Extraction Kit; Thermo Fisher Scientific). After purification, PCR products were cloned into pJET1.2/blunt (CloneJET PCR Cloning Kit; Thermo Fisher Scientific). Minipreps were prepared (GeneJET Plasmid Miniprep Kit; Thermo Fisher Scientific), and inserts were verified by sequencing.

Spleen tissue and BM were prepared as described earlier, followed by the respective FACS staining. Spleen cells were additionally seeded with 2 × 10⁶ cells per ml in 24-well plates (Greiner) and incubated with 5 μl of staurosporine/DMSO (100 nM; Abcam), 5 μl of DMSO, and medium control for 4 h at 37°C and 5% CO₂. Furthermore, spleen cells were cultivated with 2 × 10⁶ cells per ml in 24-well plates in RPMI 1640 medium lacking FCS over 24 h at 37°C and 5% CO_2. Cells were afterward carefully harvested and prepared for FACS staining.

Furthermore, macrophage/DC progenitor (MDP) cells were sorted from WT and CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice BM as (Lin⁻CD11c⁻MHC-II⁻CD135⁺CSF-1R⁺CD117^high; Lin⁺ = CD3⁺TCRβ⁺CD19⁺NK1.1⁺B220⁺Ter119⁺CD11b⁺CD4⁺CD8α⁺CD8β⁺Sca1⁺; Supplemental Fig. 4A) and seeded with 5 × 10³ cells per 100 μl in 96-well plates. Cells were cultured in the presence of 1:250 murine Flt3L supernatant and 10% murine GM-CSF supernatant in RPMI 1640 (10% FCS, 1% Pen/Strep) over the course of 4 d at 37°C and 5% CO₂. Cells were carefully harvested and stained after 24, 48, 72 and 96 h for FACS analysis.

BM samples were prepared as described above. pDC were generated by seeding 3 × 10⁶ BM cells per well in six well Primaria tissue culture plates (Corning) with 1:250 murine Flt3L CHO cell supernatant (Marburg) in OptiMEM (Life Technologies) containing 1% FCS, 1% penicillin/streptomycin, and 200 mM l-glutamine at 37°C with 5% CO₂. To generate myeloid DC, 5 × 10⁶ BM cells were seeded in 10 cm Δ Nunc cell culture dishes (Thermo Fisher Scientific) with 10% murine GM-CSF ×6310 supernatant (Marburg, Germany) in RPMI 1640, supplemented with 10% FCS, 1% penicillin/streptomycin, 50 mM 2-ME (Sigma), and 200 mM l-glutamine at 37°C in 5% CO₂. mDC cultures were supplemented on days 3 and 5 with fresh medium (50%) to ensure sufficient growth. Floating cells of both pDC and mDC cultures were collected on day 7 for FACS analysis. CD103⁺ DC were generated as previously described with minor modifications (36). Briefly, BM was seeded at 3 × 10⁶ cells per well with 10% murine GM-CSF supernatant and 1:250 murine Flt3L supernatant in six well tissue culture plates (Greiner) containing OptiMEM with 1% FCS, 1% penicillin/streptomycin, and 200 mM l-glutamine at 37°C in 5% CO₂. After the fifth day of culture, 1 ml of OptiMEM was added to reduce apoptosis, with floating cells being harvested on day 9 and reseeded at 0.6 × 10⁶ cells per well with media supplements listed for day 0. Floating cells were collected on day 15 for flow cytometry analysis. Cultured pDC were definded as CD45⁺CD11c^+/−CD11b^+/−B220⁺SiglecH⁺, mDC as CD45⁺Gr1⁻CD11c⁺CD11b⁺B220⁻ and CD103⁺DC as CD45⁺Gr1⁻B220⁻CD11c⁺CD11b^+/−CD103⁺ by flow cytometry.

Lung tissues were prepared and pooled from respective genotypes as described earlier. CD11c⁺ DC (CD45⁺Gr1⁻CD19⁻SiglecF⁻MHC-II⁺CD11c⁺) were sorted and seeded with 5 × 10³ cells per well in 96-well flat-bottomed tissue culture plates (Greiner). Cells were cultured with 25 μg/ml OVA protein (Hyglos) for 5 h prior to transgenic T cell addition.

Spleens as well as mesenteric and inguinal lymph nodes of transgenic OT-I and OT-II mice were prepared as described earlier, followed by the “untouch” sort (sort panel: Gr1, CD11b, MHC-II, CD49b, B220, TCRγδ, CD19, F4/80, CD45; control Panel: sort panel including CD3, CD4, CD8) of transgenic CD8⁺ CD4⁺ T cells, respectively. Sorted T cells were subsequently labeled with CFSE (CFSE, 1 μM; Molecular Probes) for 7 min at 37°C, and numbers of viable cells determined using trypan blue staining. Labeled transgenic T cells were afterward seeded into 96-well flat-bottomed tissue culture plates (5 × 10⁴ per well) containing pulsed CD11c⁺ DC in a ratio 1:10 with 5000 U/ml recombinant murine IL-2 (BioLegend). For OT-I T cell proliferation, cells were additionally cultured with 100 ng/ml LPS (Sigma-Aldrich) and 100 ng/ml R848 (Invivogen). Cells were harvested after 5 d for FACS analyses.

Assessments of phenotyping and proliferation were performed using a BD FACSCanto II, a BD FACSAria III, and a Sony Spectral Cell Analyzer SP6800 with conjugated mAbs and appropriate isotype controls (Table I). Prior to staining, samples were Fc blocked with mouse serum for 10 min at RT. Samples were subsequently stained according to the manufacturer’s protocol for 20 min at RT, washed with HBSS, and centrifuged at 400 × g for 5 min at 4°C for FACS flow analysis.

For intracellular Langerin/CD207 staining, samples were stained for surface markers and washed with PBS (400 × g, 5 min, 4°C). Fixation and permeabilization were prepared according to the eBioscience Foxp3/Transcription Factor Staining Buffer Set (eBioscience). Intracellular mAbs and isotype controls were added at the recommended concentrations and incubated for 30 min at RT. Cells were washed twice with the supplemented permeabilization buffer and analyzed by FACS flow analysis.

Cell viability was determined using SYTOX Blue (Life Technologies), SYTOX Green (eBioscience), and Annexin V PE (BioLegend) cell stains. Absolute cell counts were calculated using Sphero AccuCount Fluorescent Particles (7.3 μm; Spherotech).

Data sets were analyzed using BD FACSDiva Software v8.0.1 (ARIA III experiments), Sony Spectral Cell Analyzer SP6800 software, and FlowJo v10.0.

Mouse lungs were lavaged with 500 μl of HBSS and retained BAL fluids stored at −80°C until measurements. Samples were then thawed on ice, vortexed, and centrifuged at 2750 × g for 5 min at 4°C, with 10 μl of aliquots being transferred to Biocrates p180 kit filter plates. Further sample preparation and pretreatment were performed using Biocrates AbsoluteIDQ p180 kits (Biocrates Life Sciences AG), according to the manufacturer’s instructions (37). The p180 kit included all required calibration standards, internal standards, and quality control samples.

Targeted metabolomic analyses were performed using flow injection or liquid chromatography (LC) electrospray analysis tandem mass spectrometry (MS/MS). Acylcarnitines, lipids, and hexose were quantified using flow injection–MS/MS, and amino acid and biogenic amine contents were determined using LC–MS/MS with an Agilent 1290 Infinity LC system (Agilent) coupled to a QTrap5500 mass spectrometer (Sciex, Darmstadt) equipped with an ESI TurboIonSpray source and Analyst 1.6.2 Software (Sciex). Data analyses were performed using MetIDQ 4.8.0 software (Biocrates Life Sciences AG).

AM of CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} and CD11c-Cre-ADAR1^wt/wt control mice were collected from BAL fluid and purified by CD11b MagniSort depletion and CD11c MagniSort enrichment. Whole RNA was isolated using the RNeasy Plus Micro Kit (Qiagen) and stored at −80°C until further processing. Integrity of total RNA was assessed on the Bio-Rad Experion. Semiautomated library preparation was performed on the IPStar (Diagenode) using the TruSeq Stranded mRNA Kit (Illumina). Each library was quantified and qualified on a Bioanalyzer 2100 (High Sensitivity DNA Kit; Agilent Technologies). Equimolar amounts of libraries were pooled and quantified by digital PCR using the QuantStudio 3D Digital PCR System (Life Technologies). Onboard cluster generation using the HiSeq PE Rapid Cluster Kit v2 (Illumina) and paired-end 2 × 100 nt sequencing was performed on a HiSeq Rapid PE Flow Cell (Illumina) on the Illumina 1500 platform.

Raw Illumina reads were quality trimmed with Trimmomatic (38) and mapped to the mm10 mouse reference genome using HISAT2 (39). The read mappings were preprocessed with Opossum (40) using default parameters. Single-nucleotide polymorphisms (SNPs) were called on all samples with Platypus (41), applying the suggested parameters. For each SNP, high-quality reads with and without the SNP were counted separately (base quality and mapping quality ≥20). Additionally, genome features and annotations were assigned to each SNP based on the University of California Santa Cruz Known Gene model.

SNP analyses, presenting identified A-to-I RNA editing sites in AM, are presented in Supplemental Material 2 (https://www.uni-giessen.de/fbz/fb08/Inst/bioinformatik/Research/Supplements/ADAR) containing all A-to-G editing events detected in CD11c-Cre-ADAR1^wt/wt and CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} AM. An SNP is defined in WT cells if two or more reads with difference to the genomic sequence were detected.

Transcript-level read counts were estimated using RSEM and bowtie2 (42, 43). Statistically significantly differentially expressed genes were identified with edgeR (44) (false discovery rate < = 0.001 and |log(FoldChange)| ≥ 2). Analyses of differential gene expression are presented in Supplemental Material 1 (https://www.uni-giessen.de/fbz/fb08/Inst/bioinformatik/Research/Supplements/ADAR) containing 709 differentially regulated genes (p < 0.001 and fold change >2,8) in ADAR1-deficient AM compared with WT AM.

RNA sequencing data of mouse AM (ADAR1-deficient and WT) have been deposited in the ArrayExpress database of the European Molecular Biology Laboratory (https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-6477) under accession number E-MTAB-6477.

BAL fluid samples were collected as described above and centrifuged at 400 × g for 5 min at 4°C. Supernatants were stored at −80°C until analysis and protein concentrations were determined using Pierce BCA Protein Assay Kits (Thermo Fisher Scientific).

BAL fluid was diluted with SDS sample buffer followed by denaturation for 5 min at 96°C and protein separation in a 4–12% SDS-PAGE gradient. Protein bands were visualized by Coomassie blue staining and scanned using ChemiDoc XRS (Bio-Rad).

BAL fluid samples were prepared for H&E staining, as described earlier, with minor modifications. Cells in BAL fluid samples were transferred to glass microscope slides using a cytospin instrument (Shandon Cytospin 4; Thermo Fisher Scientific). Samples were fixed in 100% ice-cold methanol at −20°C for 10 min and were subsequently stained with H&E (45).

For histopathology, lungs were removed after ligation of the trachea to prevent alveolar collapse and immersion-fixed in formalin. Tissue samples were embedded in paraffin, sectioned at 2 μm- and stained with H&E after dewaxing in xylene and rehydration in graded ethanols. Visualization of polysaccharides or glycoproteins within AM was conducted using periodic acid–Schiff (PAS) reaction.

To identify neutral lipids in BAL fluid samples, Red Oil O (Sigma-Aldrich) staining was performed, as described previously, with minor modifications (46). Cytospin slides were prepared as described earlier and fixed for 30 min in 4% PFA (Roth). Slides were carefully rinsed with double-distilled H₂O, incubated in 60% isopropyl alcohol for 5 min, air dried, and subsequently stained with Oil Red O Solution (1.8 mg/ml) for 15 min. Slides were rinsed carefully with double-distilled H₂O, counterstained with Mayer Hemalaun, and washed for microscopic analysis.

Statistical analyses were performed using the GraphPad Prism software version 5 (GraphPad Software). The levels of significance of differences between groups were measured by one-way ANOVA and Tukey posttest for multiple comparisons or unpaired t test. A p value <0.05 was considered statistically significant.

DC-specific ADAR1-deficient C57BL/6 mice were generated by breeding CD11c-Cre mice (29) to ADAR1^{fΔ7−9/Δ7−9} floxed mice (28). Latter strain carries two loxP sites spanning exons 7–9 that encode for parts of the dsRNA binding and deaminase domains (28). CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice exhibited no apparent external abnormalities (47).

As CD11c expression varies among CD11c⁺ cells (Supplemental Fig. 2A), one might argue that the CD11c⁺ Cre/loxP system is biased in favor of cells expressing high amounts of CD11c. We therefore analyzed and quantified the Cre-mediated depletion of ADAR1 exon 7 to 9 on mRNA/cDNA level in the respective cell subsets. ADAR1 fragment amplification was standardized to pJET1.2/blunt control plasmids harboring the respective PCR products of sorted CD11c⁻ (WT allele) or CD8⁺ CD11c⁺ DC (exon 7 to 9 depleted allele). Splenic CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} CD8⁺ CD11c⁺ DC and CD11b⁺ CD11c⁺ DC showed almost complete deletion of exon 7 to 9, whereas exon depletion in pDC reached 53.5% (25.98 ± SD) (Fig. 1A). Concurrently, CD11c⁻ CD11b⁺ cells exhibited WT ADAR1 mRNA, and only background amounts of exon depleted mRNA.

As ADAR1 deficiency is associated with general apoptosis, we questioned if CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice boasted higher apoptotic rates in CD11c⁺ cells in comparison with the CD11c-Cre-ADAR1^wt/wt WT control. Spleen cells were therefore challenged in both starvation and apoptosis assays, revealing no differences in apoptotic rates between CD11c⁺ and CD11c⁻ cells in both WT and ADAR1-deficient mice (Fig. 1B, 1C).

We next evaluated whether ADAR1 differentially affects BM-derived differentiation of the three main DC lineages. To this end, we differentiated pDC, CD11b⁺ myeloid DC, and CD103⁺ DC, descendants of the same DC lineage as CD8⁺ DC, from CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} and WT control mice BM using mFlt3-L and mGM-CSF culturing conditions (36) (Fig. 2A–C). In comparison with WT controls, CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice were unable to expand normal numbers of CD103⁺ DC (Fig. 2C), whereas CD11b⁺ myeloid DC (Fig. 2A) and pDC (Fig. 2B) expansions were unaffected. Moreover, percentages of dead cells were not elevated in CD11c⁺ ADAR1 BM cultures, indicating that ADAR1 deficiency does not promote premature death of differentiated DC (Fig. 2A–C). Collectively, these experiments indicate that ADAR1 deficiency is dispensable for BM-derived CD11b⁺ DC development but instead preferentially suppresses the expansion of CD103⁺ DC.

To verify the role of ADAR1 in BM-derived DC subset differentiation in a second independent approach, we investigated the quantity ratio of DC progenitor cells in BM of CD11c-Cre-ADAR1^wt/wt WT control and CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice (Fig. 2E). As previously defined (48, 49), BM was screened for the presence of conventional DC progenitors, which give rise to both CD8⁺/CD103⁺ (cDC1) and CD11b⁺ (cDC2) pre-DC lineages as well as the MDP and the common DC progenitor (CDP) (48, 50–52). To address the question, if the loss of ADAR1 correlates with an increase of apoptosis due to the expression of CD11c and therefore the loss of the cDC1 lineage, MDP were cultured and brought to differentiation. MDP were therefore high-purity sorted from CD11c-Cre-ADAR1^wt/wt and CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice BM and cultured in the presence of mFlt3L and mGM-CSF over the course of 4 d (Fig. 2F). FACS analysis revealed that MDP of ADAR1-depleted mice gave rise to significantly fewer cDC1 in contrast to MDP, CDP, and cDC2, in support of the previous experiment.

To assess the in vivo relevance of the BM differentiation defect in the ADAR1 depletion model, we systematically investigated CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mouse tissues. Interestingly, multiparametric flow cytometry analyses of CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice lungs revealed a dramatic loss of CD11c⁺ SiglecF⁺ AM and moderate reductions of pulmonary CD11c⁺ MHC-II⁺ SiglecF⁻ NK1.1⁻ DC compartments (Fig. 3A, 3B). To investigate causes of moderate reductions in lung DC numbers, we subdivided lung DC into CD11b⁺ DC, CD103⁺ DC, and pDC subsets using flow cytometry (Fig. 3A, 3C). These analyses revealed an almost complete and exclusive loss of CD103⁺ DC in comparison with other DC subsets (Fig. 3D). CD11b⁺ DC exhibited impaired MHC-II expression and increased CD274 inhibitory PD-L1 expression, suggesting that these DC may display impaired T cell stimulatory activity (Supplemental Fig. 1A).

To investigate the possibility that a CD103 marker deprivation contributed to the loss of CD103⁺ DC, we extended our analyses and quantitated CD24 and CD205 expression, which are coexpressed by Batf3-dependent CD103⁺ DC (53). These experiments confirmed the absence of CD103⁺ DC in the lungs of CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice (Supplemental Fig. 1B). To eliminate the possibility of CD103⁺ DC migration, we quantitated CD103⁺ DC in draining mediastinal lymph nodes and observed a similar loss of CD103⁺ DC, whereas pDC were unaffected (Fig. 3E).

CD103⁺ DC have a pivotal role in pulmonary host defense, as they represent the most potent cross-presenting cells in lung tissue, preventing, e.g., viral immune evasion strategies. As ADAR1 deficiency in CD11c⁺ cells promoted a marked loss of CD103⁺ DCs and moderate losses of CD11b⁺ DCs, we investigated the ensuing effects on MHC-I– and -II–dependent Ag-presenting capabilities of CD11c⁺ cells. In these analyses, FACS sorted lung CD11c⁺ DC from CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} and monitored the expansion of purified CFSE-labeled OVA TCR transgenic CD4⁺ and CD8⁺ T cells as Ag-specific responder cells. OVA protein pulsed CD11c⁺ DC from CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice failed to expand normal numbers of OT-I T cells (Fig. 3F), whereas expansion of OT-II T cells was not affected. These functional experiments indicated that ADAR1 regulates the capacity to cross-present Ags through its impact on the expansion of the CD103⁺ DC lineage and confirmed absence of functional CD103⁺ DC. Taken together, these data suggest that ADAR1 is essential for lung CD103⁺ DC and AM.

The selective loss of pulmonary CD103⁺ DC prompted us to systematically investigate major DC subsets in different lymphoid and nonlymphoid organs. Analysis of DC subsets in the intestine revealed marked losses of CD103⁺ CD11b⁺ and CD103⁺ CD11b⁻ DC subsets (Fig. 4A, 4B), whereas CD103⁻ CD11b⁺ DC numbers were only moderately reduced (Supplemental Fig. 2B). Similarly, almost no CD103⁺ DC were found in liver tissues, whereas CD11b⁺ DC numbers were reduced (Fig. 4B). In contrast with all other tissues, analysis of skin tissues from CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice revealed normal numbers of CD207⁺ (langerin) CD103⁺ DC and a significant reduction of all CD103⁻ subsets (Fig. 4C).

Among lymphoid organs, spleen presented a significant decrease of CD8⁺ CD11b⁻ DC and deficiencies of CD8⁻ CD11b⁺ DC (Fig. 4D). Similarly, CD8⁺/CD11c⁺ DCs were depleted in the thymus, with CD8⁻/CD11c⁺ DCs being moderately reduced (Fig. 4E). Moreover, absolute CD45⁺ cell numbers were similar in skin, spleen, and thymus tissues from WT and knockout (KO) mice, whereas moderately increased in livers and moderately reduced in intestines of KO mice compared to the numbers in CD11c-Cre-ADAR1^wt/wt mice (Supplemental Fig. 2C). Frequencies of NK in lungs and spleens as well as pDC numbers analyzed in intestine, liver, spleen, lymph node, and thymus, did not show significant differences between the groups (Fig. 4A–D, Supplemental Fig. 2D).

In some cases, phenotypes manifest in an age-dependent manner, first occurring after weaning and or dwindling with increasing age. To exclude the possibility of a temporary, age-dependent, or delayed DC progenitor differentiation loss of CD8⁺/CD103⁺ DC in CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice, we systematically analyzed mice ranging from 8 to 62 wk of age. These analyses confirmed that CD8⁺/CD103⁺ DC were consistently depleted in lungs, spleens, and thymuses of CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice (Supplemental Fig. 2E).

In summary, these experiments suggest that ADAR1 is essential for all CD11c⁺ CD103⁺ migratory and CD8⁺ lymphoid-related DC lineages apart from langerin⁺ CD103⁺ DC in skin (Fig. 4C).

With AM numbers being drastically reduced in CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} lungs, we further investigated AM morphology and composition. To this end, BAL fluid from CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice and CD11c-Cre-ADAR1^wt/wt controls were used for cytospin preparations and revealed numerous lipoproteinaceous aggregates together with foamy AM containing massive lipid droplet accumulations, as visualized by Oil Red O staining (Fig. 5A). Further histopathologic analyses confirmed the presence of PAS-positive giant AM in alveolar spaces, indicating accumulation of carbohydrate macromolecules such as glycoproteins and glycolipids in CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice (Fig. 5B, 5C). Additionally, BAL fluid analysis indicated protein accumulation in CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice (Fig. 5D, 5E). With respect to surface phenotype, resident AM expressed elevated levels of CD274 (Fig. 5F), similar to lung CD11b⁺ DC (Supplemental Fig. 1A), indicating a regulatory phenotype.

We next characterized the composition of the lipoproteinaceous aggregates using MS-based metabolite phenotyping of BAL fluids. CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} BAL samples presented significant accumulations of various glycerophospholipids and sphingolipids, including the main surfactant components lysoPC C16:1 and SM C16:0 (Fig. 5H).

To verify the importance of ADAR1 for AM, we generated a second ADAR1-deficient mouse model under the control of the LysM locus (LysM-Cre-ADAR1^{fΔ7−9/Δ7−9)} (30). The ensuing homozygous LysM-Cre-ADAR1^{fΔ7−9/Δ7−9} mice also exhibited significantly reduced AM numbers, giant AM, and significant alveolar lipid accumulation in Oil Red O staining analysis (Fig. 5G, 5I). These data collectively indicate that AM ADAR1 deficiency disturbs alveolar surfactant catabolism, resembling pulmonary alveolar proteinosis.

As combined phenotypes of alveolar proteinosis and CD103⁺ DC deficiency in CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} animals resemble GM-CSFR deficiencies (54, 55), we quantitated GM-CSFRα (CSF2RA, CD116) and β (CSF2RB; CD131) expression on AM and leukocyte subsets. These experiments revealed selective and significant increases in CSF2RA (CD116) and CSFR2B (CD131) expression on AM, ruling out GM-CSFR loss but suggesting abnormalities in GM-CSFR circuits in ADAR1-deficient AM (Supplemental Fig. 3).

To identify molecular candidates of ADAR1-mediated RNA editing, we performed next-generation sequencing (NGS) analysis with RNA isolated from CD11b-depleted BAL cells of CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} or CD11c-Cre-ADAR1^wt/wt control mice. RNA samples were subsequently used for library preparation and sequencing followed by sequence comparison with reference genomes. This allows the identification of changes in mRNA expression as well as ADAR1-mediated editing events in AM from both genotypes.

We observed 709 differentially regulated genes (p < 0.001 and fold change >2,8) in ADAR1-deficient AM compared with WT AM (Supplemental Material 1, file name: Supplement_up_down.xls), and according to published data, ADAR1 deficiency strongly induced antiviral inflammatory and IFN responses (Fig. 6A).

However, because other CD11c⁺ cells, such as pDCs, were not affected by ADAR1 deficiency (Fig. 4A–D), we hypothesized that loss of ADAR1 editing of certain genes in AM (and CD103⁺ DC) influenced cellular development and/or homeostasis. Analyzing RNA-DNA differences (56) of NGS-derived RNA sequences from ADAR1-deficient and WT AM (Supplemental Material 2, file name SNPs_count_annotation.xlsx) confirmed the dramatic loss of A-to-I RNA editing in ADAR1-deficient AM. Examples of editing events in exonic and CDS regions of several genes are depicted in Fig. 6B. Interestingly, coding genes (e.g., Pwwp2a, Cdk13, Rpap1, and Esrra) as well as long noncoding RNAs (6820431F20Rik, Pisd-ps1, A730017L22Rik, 9330159M07Rik) are affected and demonstrate a pronounced layer of complexity in A-to-I RNA editing. Interestingly, known genes in AM development, such as GM-CSF, GM-CSFR, Bach2, or peroxisome proliferator–activated receptor γ (PPARγ) (57) were not identified as ADAR1 targets in the RDD analysis. Only one gene, estrogen-related receptor α (esrra, ERRα), shows some relation to this pathway because it acts downstream of the PPARγ coactivator-1α (PGC-1α) (58).

In this study, we could demonstrate that the conditional in vivo ablation of ADAR1 in CD11c⁺ cells has a dominant effect on CD8⁺/CD103⁺ DCs and AM in terms of differentiation, functionality, and survival.

Numerous studies have shown ADAR1 depletion ultimately results in apoptosis in an “all-or-nothing” fashion within their respective cell of interest (8, 9, 59–61). In our study, we have revealed evidence for DC subset–specific effects of ADAR1 on discrete DC populations, as general apoptosis could not explain the selective loss of the CD103⁺/CD8⁺ DC compartment.

As an alternative to apoptosis, we investigated the differentiation capabilities of discrete DC populations from BM of ADAR1-deficient mice. Here, we were able to show that CD103⁺ DC expansion is particularly dependent on ADAR1, whereas CD11b⁺ DC and pDC differentiation was unaffected without an increase in apoptotic rates. We were able to replicate these findings in a separate approach, confirming the dependency of the cDC1 (CD103⁺/CD8⁺ DC) lineage on ADAR1. Though an increase in apoptotic rates could not be detected, it remains to be tested if the observed phenotype in CD11c-Cre-ADAR1^{Δ7−9/Δ7−9} mice can be rescued by the additional deletion of MDA5 and RIG-I to fully rule out the possibility of an apoptosis-induced phenotype. However, this would exceed the scope of the current study and will be the subject of future projects.

Initial flow cytometry analysis of CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice revealed a complete loss of the lung CD103⁺ DC population as well as the presence of foamy AM. Further organ-specific dissection revealed a preferential systemic loss of the CD8⁺/CD103⁺ DC lineage in lymphoid and nonlymphoid tissues, respectively. Simultaneously, AM were dysfunctional, resembling alveolar proteinosis, which could be confirmed in a separate in vivo LysM-Cre-ADAR1^{fΔ7−9/Δ7−9} model.

In a recent study, Marcu-Malina et al. (59) showed that ADAR1 is essential for late-stage B cell lymphopoiesis but interferes with the survival of newly formed B cells. Likewise, another study reported that ADAR1 ablation decreases the formation of osteoblast progenitors, whereas osteoclast differentiation was unaffected (62). This suggest that apart from the universal masking of self-RNAs, ADAR1s’ capabilities extend to very specifically influencing the differentiation balance between subsets.

Previous studies observed a similar loss of the CD8⁺/CD103⁺ lineage and an increase in spleen weight (54, 63, 64; Supplemental Fig. 2F) in response to the loss of lipid-activated transcription factor PPARγ or the hematopoietic growth factor receptor GM-CSFR. In contrast to our findings, however, these studies observed a general decrease in migratory lymphoid CD8⁺ and nonlymphoid CD103⁺ DCs numbers in all screened tissues, including the dermal cross-presenting CD207⁺ CD103⁺ DC population (65). As both CD103⁺ DC and AM are dependent on GM-CSF, it would be a promising candidate to explain the loss of these two cell populations. In support of this, AM of CD11c-Cre-ADAR1^{fΔ7−9/Δ7−9} mice expressed elevated levels of CSF2Ra (CD116) and CSF2Rb (CD131). Initial analysis of human and murine A-to-I RNA editing sequence datasets and databases (3, 9, 66–68) showed that although murine CSF2Ra and CSF2Rb contained no A-to-I editing sites, human CSF2Rb harbored five intronic editing sites.

Though murine GM-CSFR subunits, GM-CSF, Bach2, or PPARγ were not identified among ADAR1-edited candidates, we detected the loss of ADAR1-dependent editing sites in Essra (or ERRα), a downstream target of PGC-1α (58). Interestingly, ERRα regulates inflammatory macrophage responses (69) with PGC-1α participating in repressing foam cell formation (70): two processes disturbed in ADAR1-deficient AM. Further studies will be necessary to dissect the role of estrogen-related receptors in the development and homeostasis in AM. In addition, other identified targets (coding and noncoding genes) may prove to be important in the differentiation of AM as well as in CD103⁺/CD8⁺ DCs.

Overall, the identification of A-to-I edited genes and lymph node cell RNAs has provided us with an excellent molecular tool to investigate the role of ADAR1 in AM and DC immunobiology in future studies.

We thank Miya Higuchi and Peter Seeburg, Max Planck Institute for Medical Research, Heidelberg and Irmgard Förster, Limes Institute, Bonn, for providing ADAR1-flox and LysM-Cre mice, respectively.

The authors have no financial conflicts of interest.