T cells produce interferon alpha in the TREX1 D18N model of lupus-like autoimmunity

Abstract

Autoimmunity can result when cells fail to properly dispose of DNA. Mutations in the Three-prime Repair EXonuclease 1 (TREX1) cause a spectrum of human autoimmune diseases resembling Systemic Lupus Erythematosus (SLE). The cytosolic double-stranded DNA sensor, cylic GMP-AMP synthase (cGAS) and the STimulator of INterferon Genes (STING) are required for pathogenesis, but specific cells in which DNA sensing and subsequent type-I interferon (IFN-I) production occur remain elusive. In this study, we demonstrate that TREX1 D18N catalytic deficiency causes dysregulated IFN-I signaling and autoimmunity in mice. Moreover, we show that bone marrow-derived cells drive this process. We identify both innate immune and surprisingly, activated T cells as sources of pathological IFN-α production. These findings demonstrate that TREX1 enzymatic activity is crucial to prevent inappropriate DNA-sensing and IFN-I production in immune cells, including normally low-level IFN-α-producing cells. These results expand our understanding of DNA sensing and innate immunity in T cells, and may have relevance to the pathogenesis of human disease caused by TREX1 mutation.

Introduction

Compartmentalization and disposal of polynucleotides is critical to prevent autoimmunity in metazoans. Nucleic acids represent a conserved pathogen-associated molecular pattern in viruses and bacteria, and cells have evolved pattern recognition receptors (PRRs) to detect the inappropriate presence of these macromolecules in the intra- and extracellular spaces (1). PRRs act as nucleic acid sensors to stimulate production of type I interferons (IFN-I), a family of cytokines that exert pro-inflammatory and anti-viral effects. These nucleic acid sensors detect both pathogen and self-derived polynucleotides, so cells have evolved mechanisms to precisely manage and dispose of self-genetic material to prevent the aberrant activation of nucleic acid sensors and the consequent development of auto-inflammation. Mutations in TREX1, SAMHD1, RNaseH2, ADAR1, and MDA5, genes encoding nucleic acid-processing or sensing proteins, cause human autoimmune diseases involving inappropriate IFN-I signaling (2–4). These disorders exhibit clinical overlap with the autoimmune disease Systemic Lupus Erythematosus (SLE), including the development of an interferon-stimulated gene (ISG) signature in peripheral blood, anti-nuclear autoantibodies, and inflammation of the skin. Inappropriate IFN-I activity is central to pathogenesis in these autoimmune disorders, leading to their classification as “type I interferonopathies” (5).

Mutations in the TREX1 DNA exonuclease cause a spectrum of human interferonopathies dependent upon the specific mutant allele (6). Frameshift mutations at positions within the TREX1 gene encoding amino acids 1–242, comprising the N-terminal catalytic core, cause the severe neuroinflammatory disorder Aicardi-Goutières Syndrome (AGS), likely resulting from the complete absence of TREX1 protein (7). In contrast, the missense mutation TREX1 D18N results in a catalytically inactive TREX1 protein and causes Familial Chilblain Lupus (FCL) (8). Frameshift mutations positioned between the catalytic core and the C-terminal region that controls cellular localization to the endoplasmic reticulum, cause Retinal Vasculopathy with Cerebral Leukodystrophy (RVCL) (9). Additionally, some TREX1 mutations have been associated with the more prevalent SLE, highlighting the importance of proper DNA degradation for the prevention of systemic autoimmunity (8). The varied disease phenotypes caused by TREX1 mutation suggest that TREX1 might contribute to the suppression of autoimmunity through multiple mechanisms.

We have previously demonstrated that TREX1 is a potent dsDNA exonuclease (10–12), but its precise in vivo substrate(s) have not been delineated. The biochemical activity of TREX1 suggests that dsDNA of genomic origin is a likely substrate, but other substrates, including single-stranded DNA replication intermediates and retroelements, have been proposed (13, 14). Autoimmunity in TREX1 knockout mice requires the cytoplasmic dsDNA sensor cyclic GMP-AMP synthase (cGAS), providing additional evidence that dsDNA is a primary TREX1 substrate (15, 16). Binding of unprocessed dsDNA to cGAS activates the synthase to generate the cyclic dinucleotide cGAMP, which binds to and activates the Stimulator of IFN Genes (STING) protein. Active STING recruits TBK1, which phosphorylates the transcription factor IRF3, leading to its nuclear translocation and induction of IFN-I genes (17, 18). Identification of the specific cell populations that sense undegraded TREX1 substrates through cGAS-STING and respond by producing IFN-I could facilitate the development of more effective therapies for human type I interferonopathies.

Mouse models of TREX1-mediated interferonopathies recapitulate aspects of the genotype-phenotype relationship observed with TREX1 mutant alleles in human disease. The TREX1 knockout mice develop aggressive disease and rapid mortality, mirroring the severe AGS phenotype in humans (19). In contrast, mice expressing the catalytically inactive TREX1 D18N enzyme develop milder disease mimicking FCL (20). Mice expressing TREX1 C-terminal truncation mutants develop elevated autoantibodies but no additional indications of autoimmunity, mimicking RVCL (21).

Here, we demonstrate that mice expressing the catalytically inactive TREX1 D18N allele develop a lupus-like autoimmune phenotype associated with robust induction of the IFN-I system. This finding directly links TREX1 nucleic acid catabolism to DNA-sensing, IFN-I production, and autoimmunity. We show that TREX1 inactivity within bone marrow-derived cells drives the immunological features of this phenotype, including lymphocyte activation, differentiation, and autoantibody production. Importantly, we demonstrate that TREX1 catalytic inactivity induces hematopoietic overexpression of IFN-α. Furthermore, we identify both innate immune cells and T cells as sources of this IFN-α. We show that TREX1 D18N T cells possess a functional STING signaling axis which is chronically activated, and that they produce IFN-α protein following T cell receptor (TCR) stimulation. Our work suggests that chronic IFN-α production in TREX1 exonuclease-deficient T cells acts as an inflammatory signal to facilitate expansion of autoreactive clones, leading to autoimmunity.

Materials and Methods

Mice

TREX1 D18N mice were generated as described (20). Immunophenotyping and bone marrow transplant experiments utilized 6–8 week old female TREX1 WT and D18N mice on a pure 129S1/Sv background. IFNAR^−/− mice on the C57BL/6J background were purchased from Jackson Laboratories and crossed with TREX1 D18N animals to generate F1 hybrid TREX1 D18N IFNAR^−/− mice. Pure C57BL/6J TREX1 D18N mice develop an autoimmune phenotype indistinguishable from 129S1/Sv TREX1 D18N mice (data not shown). Mixed sex TREX1 D18N STING^−/− mice on the C57BL/6J background were utilized at 8–12 weeks of age. All other experiments utilized 8–12 week old mixed sex animals on the pure 129S1/Sv background. All experiments were performed in accordance with the guidelines set forth by the Institutional Animal Care and Use Committee at Wake Forest Baptist Medical Center and the University of Virginia.

RNAseq

Total RNA was collected from splenocytes of experimental animals using the RNeasy RNA isolation kit (Qiagen), according to the manufacturer’s protocol. RNA quality was determined using an Agilent 2100 Bioanalyzer, and samples with a RIN Score >8 were utilized. RNA quality and concentration was determined using a Qubit fluorimeter, and 4 μg of total DNase-treated RNA was used for cDNA library generation using the TruSeq Stranded Total RNA LT Sample Prep Kit (Illumina, CA). Purified mRNA was fragmented, converted to cDNA, A-tailed and indexing adapters ligated, PCR amplified, purified with Ampure XP beads, and assessed again for quality using the Agilent 2100 Bioanalyzer. Samples were normalized, pooled, and run on the Illumina HiSeq 2500 using SBS v3 reagents. Collected reads were pseudo-aligned to the Ensembl mouse transcriptome (GRCm38.p6) (22) using Kallisto (23). Differential expression of aligned transcripts was measured using the DESeq2 R package (24). Gene ontology analysis was performed using the Database for Annotation, Visualization, and Integrated Discovering (DAVID) (25).

Generation of single-cell suspensions and flow cytometry

Spleens were harvested from experimental animals and processed into a single-cell suspension by mashing on a wire screen in RPMI 1640 + 10% Fetal Bovine Serum (FBS) with a 5 mL syringe plunger. Red blood cells (RBCs) were lysed using ammonium-chloride-potassium (ACK) lysing buffer (Lonza), and remaining splenocytes were washed by dilution with RPMI 1640 + 1% FBS, passed through a 70 μm nylon filter, and counted. For each stain, 10⁶ cells were incubated with the fixable viability dye Zombie Red (BioLegend) diluted 1:200 in phosphate buffered saline (PBS) for 15 minutes at room temperature.

For innate immune cell enumeration, cells were first stained with a mixture of biotinylated antibodies against CD3 (17A2, 1:50), CD19 (6D5, 1:200), CD49b (DX5, 1:200), and TER-119 (1:200) for 30 minutes on ice, then washed. Cells were then stained with anti-B220 (RA3–6B2, APC/Fire750, 1:100), CD11b (M1/70, e450, 1:400, eBioscience), CD11c (N418, PE/Cy7, 1:400), Ly6C (HK1.4, APC, 1:400), Ly6G (1A8, BV510, 1:400), CD86 (GL1, FITC, 1:100), and Streptavidin-PE in FACS buffer (PBS + 2% FBS) for 30 minutes on ice. For T cell enumeration, cells were stained with anti-CD3 (17A2, AF700, 1:50, eBioscience), CD90.2 (30-H12, PerCP/Cy5.5, 1:200), CD4 (RM4–4, APC/Fire 750, 1:200), CD8 (53–6.7, V500, 1:200, BD Biosciences), CD44 (IM7, BV421, 1:400), CD62L (MEL-14, FITC, 1:400), CD69 (H1.2F3, PE, 1:100), and SCA-1 (D7, PE/Cy7, 1:250). For B cell enumeration, cells were stained with anti-CD19 (6D5, PerCP/Cy5.5, 1:200), B220 (as above), CD138 (281–2, PE, 1:200), GL7 (AF647, 1:200), CD95 (Jo2, BV510, 1:100, BD Biosciences), and CD69 (H1.2F3, FITC, 1:100). Except where otherwise specified, all antibodies were acquired from BioLegend. Cells were washed 3 times with FACS buffer after staining, fixed in a 1% paraformaldehyde (PFA) solution for 30 minutes on ice, then acquired on an LSR Fortessa X-20 (BD Biosciences). Data were analyzed using FlowJo™ analytical software (FlowJo, LLC). Gating strategies can be found in Supplemental Figure 1. Innate immune gating strategies were based on an established approach (26).

For measurement of protein phosphorylation, cells were processed as described above, then incubated for 1 hour at 2 × 10⁶ cells/mL and 37 °C with or without 10 μg/mL DMXAA (Invivogen). After harvesting and pelleting, cells were stained with a viability dye. For pTBK1 interrogation, cells were stained with surface antibodies allowing delineation of T cells and pDCs, then fixed and permeabilized using the eBioscience Foxp3 Transcription Factor Staining buffer set. Cells were then stained with anti-pTBK1 (D52C2, PE, 1:200) or an equal concentration of rabbit IgG isotype control (DA1E, PE, 1:50). Cells were washed 3 times, then acquired. For measurement of pIRF3, cells were first stained with anti-CD19 (6D5, FITC, 1:200), then fixed with 1% PFA. Cells were washed twice with PBS, resuspended in 100 μL of PBS, then slowly dripped into 900 μL of ice-cold 100% methanol while gently vortexing and stored at −20 °C overnight. The next day, cells were pelleted, resuspended in FACS buffer, and stained with anti-CD3 (PE), CD4 (BUV737), CD8 (BUV395), B220 (APC/Fire750), CD11c (PE/Cy7), and CD11b (e450) at the concentrations described above. We verified that these antigens survive fixation with PFA and methanol, and that the CD19-FITC signal does not degrade during the fixation process. The cells were then stained with anti-pIRF3 (D6O1M, 1:100) or concentration-matched anti-rabbit IgG isotype control (DA1E, 1:500), washed 3 times, stained with goat anti-rabbit polyclonal antibody (Alexafluor 647, 1:1000, Abcam), then washed 3 times and acquired. All phospho-antibodies and isotype controls were from Cell Signaling.

Bone marrow transplant

Bone marrow transplants were conducted as 3 independent experiments with 3–5 mice per group in each experiment. Recipient animals were female 6–8 week old WT and D18N mice. In preparation for the transplant, animals were maintained on filtered antibiotic water containing 800 mg/L sulfamethoxazole and 160 mg/L trimethoprim (Sigma-Aldrich) for 3 days before irradiation. On the day of the transplant, bone marrow was harvested from female 6–8 week old WT and D18N donor mice by flushing the femur and tibia with RPMI 1640 + 1% FBS. RBCs were lysed using ACK buffer, and remaining cells were resuspended in RPMI 1640 + 10% FBS and counted. T cells were depleted from isolated bone marrow using a CD3ε MicroBead magnetic depletion kit and LS column (Miltenyi Biotec), in order to prevent transfer of autoreactive T cell clones. Successful depletion of T cells was confirmed by staining cells with anti-mouse CD3-PE (1:50, BioLegend) and CD90.2-APC (1:200, BioLegend), and verifying the absence of positive cells by flow cytometry. Depleted cells were resuspended in serum-free RPMI 1640 for injection. Recipient animals were administered a single sub-lethal dose of gamma radiation equaling 900 rads using a cesium source. A total of 5 × 10⁶ T cell-depleted bone marrow cells of either genotype were injected in a volume of 500 μL through the tail vein into irradiated animals. Animals were bled at 4 week intervals following the transplant, and euthanized 15 weeks after the transplant procedure.

Successful bone marrow engraftment was assessed by collecting Peripheral Blood Mononuclear Cells (PBMCs) and employing a qPCR-based genotyping strategy. In brief, blood was collected from experimental animals 12 weeks after transplant by facial vein puncture. A total of 5–6 drops of blood were collected into tubes containing 1 mL 4% (w/v) sodium citrate in water, 1 mL of RPMI 1640 + 1% FBS was added, and the tubes mixed by inversion. This solution was underlaid with 1 mL of Histopaque 1077 (Sigma-Aldrich), and the mixture centrifuged at 400 x g for 20 minutes. PBMCs were aspirated from the resulting interface and washed twice by pelleting and resuspension in RPMI 1640 + 1% FBS. DNA was isolated from purified PBMCs using the DNeasy DNA isolation kit (Qiagen) according to the manufacturer’s protocol, and the quality and concentration of DNA assessed using a Nanodrop spectrophotometer.

The TREX1 genotype of engrafted hematopoietic cells was assessed using qPCR (Figure 2A). The total concentration of DNA in each sample was quantitated using a Taqman primer/spanning a single exon of ISG15 (Thermofischer Scientific) and a standard curve of genomic DNA of known concentrations across 4 logs, while the concentration of TREX1^WT/WT DNA was determined using a primer/probe set specifically recognizing the TREX1 WT allele and a similar standard curve of genomic DNA (Forward: 5’-CCC ATC TCC TCC CCA GGC-3’, Reverse: 5’-GGC CAG TGG CTT CCA GGT C-3’, Probe: 5’-[Fluorescein (FAM)]-CCC ATG GTC ACA TGC AGA CCC TCA TCT TC-[Black Hole Quencher 1 (BHQ1®)]-3’). Each reaction consisted of 5 μL of 2X Taqman gene expression master mix (Thermofischer Scientific), 0.5 μL of 20X primer/probe mix (18 μM primer, 5 μM probe), and 5 μL of template DNA. Data were collected on a 7500 Real-Time qPCR system (Applied Biosystems) and analyzed using the 7500 analysis software. The ratio of TREX1 WT DNA concentration::total DNA concentration was used as a measure of successful engraftment.

Figure 2. TREX1 catalytic inactivity in hematopoietic cells induces IFN-I signaling and lymphocyte activation.

(A) Experimental setup and measurement of hematopoietic TREX1 genotype for bone marrow transplant experiments. Legend denotes donor genotype→recipient genotype. Genotype (right panel) represents the concentration of TREX1 WT DNA vs total DNA collected from PBMCs (B) Splenocyte and PBMC ISG expression at time of sacrifice. ISG score represents averaged fold induction of IFIT1, IFIT3, and Usp18. Untreated animals included as reference point (not included in statistical analysis) (C) Total cell counts and (D) CD86 expression of splenic innate immune cells. (E) Relative SCA-1 expression on splenic CD4⁺ or CD8⁺ T cells, normalized to WT→WT MFI, and (F) percentage of splenic effector or memory CD4 and CD8 T cells that are CD69⁺. (G) Percentage of peripheral CD4⁺ T cells that are effector or memory over time. (H) Total, activated, germinal center, and plasma B cell counts, and (I) serum α-dsDNA autoantibody levels. All figures represent 2–3 independent experiments with a total of 5–7 female mice. Mice were age-matched to those in Figure 1. Error bars represent SEM. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001 by one-way ANOVA with Dunnett’s post-hoc test against WT→WT control group.

ELISAs

Serum was collected from whole blood incubated at room temperature for 1 hour and centrifuged at 1500 x g for 15 minutes at 4 °C. For measurement of anti-dsDNA autoantibodies, serum was diluted 1:2000 with the supplied diluent and anti-dsDNA autoantibody titer measured using a commercial kit (Alpha Diagnostics), according to the manufacturer’s protocol. For measurement of IFN-α, serum or cell culture supernatants were diluted 1:125 (LCMV samples) or 1:1 (all others) with supplied diluent and IFN-α concentration measured using a commercial kit (Verikine high-sensitivity IFN-α all subtype ELISA kit, PBL Assay Science).

qRT-PCR measurement of IFN-I and ISGs

RNA was collected from splenocytes or purified cells as described above. On-column DNase-treatment was performed during RNA isolation using a DNase kit (Qiagen), according to the manufacturer’s supplemental protocols. Quality and concentration of RNA was assessed using a Nanodrop spectrophotometer. A maximum of 1 μg of RNA was converted to cDNA using the Protoscript II cDNA conversion kit (Applied Biosystems), according to the manufacturer’s supplied protocol. The cDNA product was diluted 5X with nanopure water and reactions performed as described above. ISG, IFN-β, and RPLP0 Taqman primer/probe mixes were from Thermofisher Scientific. The IFN-α assay consisted of 6 forward primers, listed 5’→3’ (ATA CTT CCA CAG GAT CAC TGT GTA CCT G, ATA TTT CCA CAG GAT CAC TGT GTA CCT G, ATA CTT CCA CAG CAT CAC TGT GTT CCT G, ATA CTT CCA CAG CAT CAC TGT GTA CCT G, ATA CTT CCA CAG GAT CAC TGT TTA CCT G, ATA CTT CCA CAG GAT CAC TGT GTT CCT G), 4 reverse primers (GGC TCT CCA GAC TTC TGC TCT GA, GGT TCT CTG GAC TTC TGC TCT GA, GGC TCT CCA GAT TTC TGC TCT GA, GGT TCT CCA GAC TTC TGC TCT GA), and a 5’-FAM, 3’-BHQ1-labeled hydrolysis probe ([FAM]-AGA AGA AAC ACA GCC CCT GTG CCT GG-[BHQ1]), all from Eurofin Genomics. Forward primers were used at 150 nM, reverse primers at 225 nM, and the probe at 250 nM. Data were collected on a 7500 Real-Time qPCR system (Applied Biosystems), and analyzed using the accompanying 7500 analysis software. Fold expression measurements were made using the ΔΔCT method, with normalization against the house-keeping gene RPLP0.

For pDC IFN-I expression measurements, low cell inputs necessitated concentration of cDNA products before downstream qRT-PCR analysis. RNA was collected from a total of 1 × 10⁵ purified pDCs. Following reverse transcription, a 1:10 volume of 3 M sodium acetate was added, followed by glycogen to a final concentration of 0.25 μg/μL. This solution was diluted with 4 volumes of ice-cold 100% ethanol and incubated at −80 °C for 1 hour. The solution was then centrifuged at 16000xg for 30 minutes and carefully decanted, avoiding the glycogen pellet. The pellet was washed twice by adding 500 μL ice-cold 70% ethanol, centrifuging for 10 minutes, and decanting. The pellet was then air dried, and resuspended in 30 μL of nuclease-free water for downstream analysis. An equal number of splenocytes and T cells were subjected to this protocol in tandem with pDC samples to ensure that precipitation did not introduce any artifacts, which yielded similar results to larger cell inputs (data not shown).

Purification of innate immune, T, and B cells

For each separation, splenocyte single cell suspensions were stained on ice for 10 minutes in FACS buffer + 2 mM EDTA containing a cocktails of biotinylated antibodies. For innate immune negative selection, cells were stained with the same biotinylated antibody cocktail used for innate immune cell surface staining, described above. For T cell negative selection, splenocytes were stained with anti-CLEC9a (7H11, 1:50, Miltenyi Biotec), CD19 (6D5, 1:200), CD138 (281–2, 1:200), Ly6G (1A8, 1:400), TER-119 (1:220), CD11c (N418, 1:200), and CD11b (M1/70, 1:400). For B cell negative selection, cells were stained with anti-CD3 (17A2, 1:50), CD49b (DX5, 1:200), CD115 (AFS98, 1:200), and CLEC9a, Ly6G, and TER-119 as above. Stained and unstained cells were separated using Anti-Biotin Microbeads and an LS column, according to the manufacturer’s protocol (Miltenyi Biotec). B cells underwent a secondary round of purification, using a positive selection against CD19 and CD138, as above. Except where otherwise specified, biotinylated antibodies were from BioLegend. Purity of cell preparations was assessed after every separation using flow cytometry and similar panels to those described above (Figure S3C–F).

Cell sorting

To enrich for T cells before sorting, splenocyte single cell suspensions were stained on ice for 10 minutes in FACS buffer + 2 mM EDTA containing anti-CD19 (1:200), TER-119 (1:200), Ly6G (1:400) biotinylated antibodies, then washed. Stained and unstained cells were separated using Anti-Biotin Microbeads and an LS column, according to the manufacturer’s protocol (Miltenyi Biotec). For each sorting sample, enriched T cells from 2 mice were pooled to increase sorting yield. Cells were then stained with a T cell-specific panel of fluorochrome-conjugated antibodies similar to that described above. CD3⁺ CD4⁺ or CD8⁺ cells were sorted based on CD44 expression at 4 °C on a Beckman Coulter MoFlo Astrios EQ using a 70 μm nozzle, and immediately pelleted and lysed for subsequent RNA isolation and qRT-PCR analysis, as described above. Purity of all sorted populations was >95%.

For sorting of pDCs, pooled splenocyte suspensions were generated from 3 mice for each sample. Magnetic depletion of lymphocytes and neutrophils was accomplished as described above, with the addition of biotinylated anti-CD3 (1:50). Cells were stained with viability dye and anti-CD90.2, CD4, CD8, B220, Ly6C, and CD11c. Contaminating T cells were excluded as CD90.2⁺ CD4⁺ or CD8⁺ events, while contaminating B cells were excluded as B220⁺ Ly6C⁻ events. pDCs were sorted as B220⁺ Ly6C⁺ CD11c^int cells using a Becton Dickinson FACSAria with a 100 μm nozzle. Purity was routinely >95%.

Viral infection

The IFN-α expression assay was validated using virally-challenged mice. Mixed-sex 8–12 week old WT mice on the 129 background were challenged with 2 × 10⁴ PFU LCMV-Arm by intraperitoneal injection. After 48 hours, animals were euthanized and splenocyte single-cell suspensions generated. A total of 5 × 10⁶ cells were set aside for RNA isolation, and the remainder of the cells subjected to innate immune cell purification as described above. RNA was isolated with on-column DNase treatment from whole splenocytes, innate cells, and non-innate cells, and subjected to IFN-α and ISG expression measurements, as described above.

Cell culture experiments

TREX1 WT and D18N T cells were purified by column purification or sorted, as described above. For transwell experiments, source cells were incubated at a concentration of 1 × 10⁶ cells/mL and 37 °C for 1 hour with or without 10 μg/mL DMXAA. In addition, TREX1 WT reporter cells were incubated under similar conditions in either media or 20 μg/mL anti-mouse IFNAR (MAR1–5A3, Leinco Technologies). Cells were then washed four times by dilution with 1 mL media, pelleting, and decanting. Cells were added to a 96-well transwell apparatus with a 0.4 μm pore size (Corning). A total of 235 μL of 1 × 10⁶ cells/mL source cell solution was added to the bottom chamber of each well, and then the upper chamber gently overlaid. A total of 80 μL of 2 × 10⁶ cells/mL reporter cell solution was added to the upper chamber of each well, and the plate covered. Cells were incubated for 24 hours at 37 °C, then reporter cells from the upper chamber collected for RNA isolation and qRT-PCR analysis, as described above.

For long-term in vitro measurement of the ISG signature, 2 × 10⁵ purified WT or D18N T cells in 100 μL media were added to a 96-well plate. This was then supplemented with 100 μL media, 40 μg/mL anti-IFNAR, 40 μg/mL mouse IgG₁ isotype control (Leinco Technologies), 2 × 10⁶ beads/mL anti-CD3/CD28-coated beads (Miltenyi Biotec), or T cell activation beads + anti-IFNAR/isotype control. For experiments utilizing sorted naïve T cells, wells were pre-coated with 5 μg/mL anti-CD3/CD28 overnight and then washed once with PBS in lieu of activation beads. Cells were incubated for 72–96 hours at 37 °C, then harvested and RNA collected for qRT-PCR analysis. For in vitro measurement of IFN-α production, 2 × 10⁵ purified WT or D18N T cells in 100 μL media, or 2.5 × 10⁴ pDCs in 50 μL media were added to a 96-well plate. T cells were supplemented with 100 μL of media, 20 μg/mL DMXAA, 2 μM CpG DNA, or 2 × 10⁶ beads/mL anti-CD3/CD28-coated beads, while pDCs were supplemented with 50 μL of media, 20 μg/mL DMXAA, or 2 μM CpG DNA. Supernatants were collected after 24 hours and IFN-α measured as described above.

Results

TREX1 catalytic inactivity drives aberrant IFN-I signaling and lupus-like autoimmunity

TREX1 catalysis-dependent and -independent mechanisms contribute to inflammation and autoimmunity in the TREX1 null mouse (13, 14, 27, 28). We utilized TREX1 D18N catalytically deficient mice to specifically address the effects of failed DNA degradation in TREX1-mediated disease and to demonstrate a definitive link between DNA sensing, IFN-I signaling, and inappropriate immune activation. We began by assessing differences in splenic gene expression between TREX1 WT and D18N mice using RNAseq. Seventeen of the nineteen most significantly and robustly upregulated genes in TREX1 D18N animals (FDR < 0.001 and fold change > 4) were identified as interferon-stimulated genes (Figure 1A). Consistent with this finding, gene ontology analysis demonstrated significant enrichment in signaling pathways associated with immune responses and response to viral infection. To identify the extent of interferon signaling caused by the TREX1 D18N mutation, we measured the expression of a subset of ISGs in tissues of TREX1 WT and D18N mice using qRT-PCR. ISG induction was readily detected in multiple tissues, indicating systemic IFN-I sensing (Figure 1B). Thus, TREX1 catalytic inactivity is sufficient to induce DNA-sensing and inappropriate ISG induction.

Figure 1. TREX1 catalytic inactivity drives aberrant IFN-I signaling and lupus-like autoimmunity.

(A) Volcano plot (left) and gene ontology analysis (upper right), generated from differential expression analysis of RNAseq data, demonstrating upregulation of ISGs and enrichment of immune signaling pathways in TREX1 D18N mice (3 mice per genotype, mixed sex, 6–8 weeks of age). (B) ISG expression within TREX1 D18N tissues (SG: salivary gland, CLN: contralateral lymph node, 3–4 mice per genotype, mixed sex, 8–12 weeks of age). (C) FACS measurement of total cell counts and (D) CD86 expression of splenic innate immune cell populations (Neu: Neutrophil, Mac: Macrophage, Mono: Monocyte, M-DC: Myeloid (CD11b⁺) dendritic cell (DC), L-DC: Lymphoid (CD11b⁻) DC, pDC: Plasmacytoid DC. 10–12 mice per genotype, mixed sex, 24 weeks of age). (E) FACS measurement of relative SCA-1 expression on splenic CD4⁺ and CD8⁺ T cells, normalized to WT MFI, (F) cell counts for splenic CD4⁺ T cell subsets (Naïve: CD62L^high CD44^low-int, Effector: CD62L^low CD44^high, Memory: CD62L^high CD44^high), and (G) percent of total peripheral CD4⁺ T cells that are effector or memory over time (6–10 mice per genotype, mixed sex, 24 weeks of age). (H) FACS measurement of total, activated, germinal center, and plasma splenic B cell counts (10–12 mice per genotype, mixed sex, 24 weeks of age). Error bars represent SEM. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001 by unpaired student’s t-test. All data is representative of 2–3 independent experiments.

We next utilized high-dimensional flow cytometry to examine cellular indications of inflammation and IFN-I signaling in TREX1 D18N mice (Figure S1A–C). TREX1 D18N animals exhibited significantly increased numbers of innate immune cells, including macrophages, monocytes, lymphoid dendritic cells, and plasmacytoid dendritic cells (pDCs) (Figure 1C). Innate immune cells from these mice also expressed higher levels of the costimulatory molecule CD86, indicative of inflammation-associated maturation (Figure 1D). In addition to changes within innate immune cells, lymphocytes were also dramatically altered by the TREX1 D18N mutation. All T cells from TREX1 D18N mice uniformly expressed high levels of SCA-1 (Figure 1E). T cells strongly upregulate SCA-1 following IFN-I exposure (29), but this protein can also be found on T memory stem cells (30, 31) and on classical memory cells (32). We confirmed that SCA-1⁺ T cells did not express other markers of T memory stem cells, and observed that SCA-1 was universally expressed by naïve D18N T cells (data not shown). Thus, this observation likely reflects exposure to IFN-I. IFN-I sensing by T cells acts as a potent inflammatory and pro-activation signal (33), and IFN-I signaling is required for T cell-mediated autoimmunity in TREX1 null animals (27). TREX1 D18N animals exhibited robust CD4⁺ T cell differentiation, similar to that reported in TREX1 knockout mice. The CD4 compartment skewed towards a more antigen-experienced phenotype in both the spleen (Figure 1F) and peripheral blood as early as ten weeks, and became increasingly pronounced with age (Figure 1G). We also observed increased numbers of total and activated B cells, increased germinal centers, and increased numbers of antibody-producing plasma cells in the spleens of TREX1 D18N animals, consistent with the autoreactive B cell response of lupus-like disease (Figure 1H). Together, these results establish a definitive link between loss of TREX1 exonuclease activity, aberrant activation of the IFN-I system, and development of cellular correlates of lupus-like autoimmunity.

TREX1 catalytic inactivity in hematopoietic cells induces lupus-like autoimmunity

In order to identify the cellular origins of IFN-I signaling and immune activation in TREX1 D18N mice, we generated bone marrow chimeric animals expressing TREX1 D18N in only hematopoietic cells, only somatic cells, or in both compartments. TREX1 WT and D18N mice were irradiated and injected with T-cell depleted bone marrow from TREX1 WT and D18N donors, and a qPCR strategy was used to confirm successful engraftment of the desired hematopoietic genotypes (Figure 2A). We first assessed immune phenotypes in experimental animals after 15 weeks by measuring ISG expression and comparing to unmanipulated TREX1 WT and D18N mice. Expression of TREX1 D18N in either compartment alone was sufficient to induce a modest ISG signature, whereas the presence of TREX1 D18N within both compartments induced a more robust response similar to unmanipulated TREX1 D18N animals (Figure 2B). This initial finding suggested that catalytic inactivity of TREX1 in both hematopoietic and non-hematopoietic cells contributes to IFN-I signaling. However, further immune cell analyses revealed increased numbers of innate immune cells only within TREX1 D18N bone marrow recipients (Figure 2C), accompanied by significantly increased expression of CD86 within these populations (Figure 2D), consistent with auto-inflammation observed in unmanipulated TREX1 D18N animals. We also observed significant induction of T cell SCA-1 expression in all mice that received TREX1 D18N bone marrow (Figure 2E), accompanied by increased CD4⁺ and CD8⁺ T cell activation (Figure 2F), increased differentiation of circulating CD4 T cells (Figure 2G), and splenic CD4 T cells (data not shown). As expected, this CD4 T cell response coincided with increased B cell activation and proliferation. TREX1 D18N bone marrow recipients exhibited increased numbers of total and activated B cells, as well as a trend towards increased germinal center and plasma cells (Figure 2H). Importantly, increased serum concentrations of anti-dsDNA autoantibodies were only observed in animals expressing TREX1 D18N in hematopoietic cells (Figure 2I). This major diagnostic criterion of SLE in humans has been previously observed by us in unmanipulated TREX1 D18N animals (20). Thus, ISG signaling results from TREX1 D18N expression in either bone marrow or somatic cells, but the cellular correlates of autoimmunity require expression only in hematopoietic cells.

TREX1 D18N innate immune and T cells express IFN-I

Having identified a molecular signature of IFN-I signaling and a hematopoietic origin of immune activation in the TREX1 D18N mice, we next sought to identify specific cell populations producing IFN-I. IFN-Is are induced by DNA sensing through the cGAS-STING signaling axis and other PRRs (34, 35), and act through the common type I IFN-α/β receptor (IFNAR) (36). IFNAR expression was required for development of lupus-like disease in TREX1 D18N mice, supporting the IFN-I and autoimmunity connection. TREX1 D18N IFNAR^−/− mice were completely rescued from autoimmune mortality, exhibited no detectible ISG overexpression, and showed no indications of increased lymphocyte activation, differentiation, or anti-nuclear autoantibody (ANA) production (Figure S2A–E). Similarly, TREX1 D18N STING^−/− mice exhibited no evidence of ISG induction, ANA production (Figure S2B, E), or accelerated mortality (data not shown), and TREX1 D18N cGAS−/− mice similarly show no evidence of autoimmunity (37). Thus, exonuclease deficiency in TREX1 D18N mice causes cGAS-STING-dependent IFN-I production which is sensed through IFNAR, leading to autoimmunity.

IFN-α has been implicated as a key molecule in the pathophysiology of SLE (38, 39) and has been found in the cerebrospinal fluid and serum of AGS patients (40, 41). As such, we sought to measure IFN-α production in TREX1 D18N animals. We were able to detect IFN-α in cells stimulated with TLR or STING agonists using an established flow cytometry protocol (42), but levels in unstimulated TREX1 D18N cells were below the limit of detection, consistent with a low-level IFN-I response (data not shown). As such, we designed a qRT-PCR assay to simultaneously measure expression of all 14 IFN-α genes. To validate this assay, IFN-α expression was measured in the spleens of WT mice infected with Lymphocytic Choriomeningitis Virus, Armstrong strain (LCMV-Arm), which potently induces IFN-α production predominantly within innate immune dendritic cells (43–45). As expected, we observed significantly increased levels of IFN-α expression in LCMV-Arm-infected splenocytes relative to uninfected animals. Further, we measured an 11-fold enrichment of IFN-α expression in purified innate immune cells from LCMV-infected splenocytes (Figure 3A, S3A–B), consistent with a predominantly innate immune origin. Thus, this assay represents an appropriate method for measuring IFN-α expression.

Figure 3. TREX1 D18N T cells spontaneously express IFN-I.

(A) IFN-α expression in uninfected WT splenocytes, LCMV⁺ splenocytes, purified LCMV⁺ non-innate immune cells, and purified LCMV⁺ innate immune cells (data represents cells from three virally-challenged mice 48 hours post-infection). (B) IFN-α gene expression in splenocytes from unchallenged TREX1 WT and D18N mice (26–30 mice per genotype). (C) Serum IFN-α levels in unchallenged TREX1 WT or D18N mice and in WT LCMV-challenged mice, as in (A). (D) IFN-α expression in whole splenocytes, or in innate immune cells, pDCs, B cells, and T cells purified from splenocytes. Normalized to WT splenocyte expression (Innate, B, and T cells represent 3 magnetic separations with 3 pooled samples from 2 mice of each genotype. pDCs were FACS sorted from pooled samples and represent 3 samples of each genotype). (E) ISG expression in whole splenocytes, T cells, and B cells from TREX1 WT and D18N mice (3 separations, 3 pooled samples from 2 mice of each genotype). (F) IFN-α expression in naïve or effector/memory CD4 or CD8 T cells sorted from the spleens of TREX1 WT and D18N mice (3 independent sorts, 3–5 pooled samples from 2 animals of each genotype). All mice were mixed sex and 8–12 weeks of age at time of sacrifice. Error bars represent SEM. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001 by unpaired student’s t-test (B-D, WT vs D18N comparisons in E, F) or one-way ANOVA with Dunnett’s post-hoc test (D18N population comparisons in E, F).

We next sought to identify active isoforms and cellular sources of IFN-I in the TREX1 D18N mouse. IFN-α and IFN-β were upregulated 5-fold in the spleens of TREX1 D18N animals, with IFN-α transcripts approximately 25-fold more abundant than IFN-β transcripts (Figure 3B, S4). Where measured, the relative expression levels of IFN-α and IFN-β were well-correlated, so we focused on the more abundant IFN-α due to its hypothesized role in the pathogenesis of lupus and lupus-like disease. Elevated levels of IFN-α protein were observed in the serum of TREX1 D18N mice (Figure 3C), confirming that it is both expressed and secreted. The absolute concentration of IFN-α was several orders of magnitude below that observed in acute LCMV-Arm infection, consistent with a low-level, “smoldering” inflammatory response caused by TREX1 inactivation.

Innate immune cells are major IFN-α producing cells, with pDCs specialized for its production (46). As such, we hypothesized that exonuclease deficient TREX1 D18N pDCs sensing undegraded cytosolic DNA might be a major source of IFN-α production. To test this, we purified a mixed population of innate immune cells from TREX1 WT and D18N spleens, resulting in an approximately 11-fold enrichment of pDCs (Figure S3A–B). Surprisingly, IFN-α expression in TREX1 D18N innate immune cells was similar to levels seen in TREX1 D18N whole splenocytes (Figure 3D), suggesting that non-innate populations also contribute to IFN-α expression. To more specifically assess the contribution of pDCs, we sorted pDCs to a purity of >95% and measured IFN-α expression. Interestingly, we could not detect elevated expression of IFN-α in TREX1 D18N pDCs (Figure 3D). This population therefore likely does not contribute to spontaneous IFN-α expression in the TREX1 D18N spleen.

We next asked if other TREX1 D18N splenic cell populations might exhibit elevated IFN-α expression. Although not traditionally regarded as major IFN-α-producing cells, T and B cells represent 70–80% of splenic leukocytes, and therefore strongly influence whole-spleen gene expression measurements. We purified TREX1 WT and D18N T and B cells to >99% and >97% purity, respectively (Figure S3C–E) and measured IFN-α induction. TREX1 D18N T cells exhibited robust enrichment of IFN-α expression relative to whole splenocytes, whereas expression was significantly reduced in B cells (Figure 3D). We also noted significantly higher ISG expression in T cells relative to both splenocytes and B cells, consistent with the observed pattern of IFN-I induction (Figure 3E). To determine if T cell IFN-α expression tracked to a specific cell population, we sorted naïve or CD44^high effector/memory (E/M) CD4 and CD8 T cells from the spleens of TREX1 WT and D18N animals, and measured induction of IFN-α. Relative to TREX1 D18N splenocytes, significant enrichment of IFN-α expression was observed only in naïve CD4 and CD8 T cells (Figure 3F). Interestingly, while IFN-α induction in TREX1 D18N E/M CD8 T cell did not enrich relative to splenocytes, expression was still elevated over WT E/M CD8s. In contrast, we observed no significant induction in TREX1 D18N E/M CD4 T cells relative to their WT counterparts. Thus, the TREX1 D18N T cell IFN-α response appears to be delocalized across various populations and differentiation states, but is most strongly enriched within naïve cells. These findings demonstrate that catalytic inactivity of TREX1 D18N within both innate immune and T cells activates DNA sensing and IFN-α/β expression.

T cells possess a functioning STING axis which is constitutively active in TREX1 D18N T cells

TREX1-mediated autoimmunity involves inappropriate T cell activation, and IFN-I can act as an inflammatory signal supporting T cell activation. As such, we elected to further explore TREX1 D18N T cell IFN-I production, as a potentially self-supplied, pro-autoimmune signal. We first sought to confirm that TREX1 D18N and WT T cells possess the capacity to respond to the presence of undegraded cytosolic DNA through a cGAS-STING-dependent mechanism. We found that all T cell subsets expressed cGAS, STING, and IRF3/IRF7, which control IFN-β and IFN-α/β transcription, respectively (47) (Figure 4A–B). Relative to whole TREX1 WT splenocytes, WT E/M and D18N naïve and E/M T cells were significantly enriched for cGAS expression. Interestingly, we observed slight enrichment of STING expression in both TREX1 WT and D18N E/M CD4, but not CD8 T cells. cGAS-STING expression levels were therefore not well-correlated with the observed pattern of TREX1 D18N T cell IFN-α expression (Figure 3F), suggesting alternative regulatory mechanisms. Similar to whole splenocytes (Figure 1A), IRF7 expression was strongly expressed within all TREX1 D18N T cell subsets. Thus, TREX1 D18N T cells express all molecular components of the DNA sensing and signaling pathways required for cytosolic DNA detection and IFN-α/β production.

Figure 4. T cells possess a functioning STING axis which is constitutively active in TREX1 D18N T cells.

(A, B) Expression of cGAS-STING signaling molecules in naïve or effector/memory CD4 or CD8 T cells sorted from the spleens of both TREX1 WT and D18N mice (3 independent sorts, 3–5 pooled samples from 2 animals of each genotype, mixed sex animals 8–12 weeks of age). (C, D) FACS measurement of TBK1 phosphorylation in splenic CD4 or CD8 T cells and pDCs with or without a 1 hr stimulation with DMXAA. (E, F) FACS measurement of IRF3 phosphorylation in the same cell populations. Panels (C-F) represent 2–3 independent experiments with a total of 5–6 mixed sex mice 8–12 weeks of age. Error bars represent SEM. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001 by one-way ANOVA with Dunnett’s post-hoc test (A, B) or unpaired student’s t-test (D, F).

We next examined the functionality and activation state of the STING pathway in TREX1 WT and D18N T cells and pDCs. We attempted to measure basal IRF7 phosphorylation in TREX1 D18N cells by western blot, but it was below the limit of detection in all animals tested (data not shown). Using flow cytometry, we were able to observe increased phosphorylation of TBK1 (Figures 4C–D) and IRF3 (Figures 4E–F) in both T cells and pDCs following stimulation with the small-molecule STING agonist 5,6-dimethylxanthenone-4-acetic acid (DMXAA) (48). Furthermore, TREX1 D18N T cells exhibited a heightened response to stimulation. This confirms that STING is functional within T cells, consistent with recent reports (49–51), and suggests that STING activation in TREX1 D18N T cells may lead to heightened cytokine responses. We noted that pDCs exhibited a smaller relative change in pTBK1/IRF3 signal following DMXAA stimulation compared to T cells, but this may have been a consequence of higher background staining. We also noted that TREX1 D18N pDCs exhibited a dampened response to DMXAA stimulation relative to WT, with potential downstream consequences for cytokine production. Importantly, we observed increased basal TBK1/IRF3 phosphorylation in TREX1 D18N T cells (Figures 4C–F), indicating spontaneous activation of the primary STING signaling transducers. Thus, TREX1 inactivation in T cells results in constitutive activation of the cGAS-STING pathway.

Activated TREX1 D18N T cells produce IFN-α protein

We next asked if this evidence of STING activation coincided with the spontaneous production of IFN-I protein. TREX1 WT or D18N “source” T cells were placed in the bottom chamber of a transwell apparatus, separated from an upper chamber by a pore-containing membrane. IFN-I produced by source cells could diffuse across this membrane to induce ISG expression in “reporter” TREX1 WT T cells placed in the upper chamber (Figure 5A). A ~2-fold upregulation of reporter ISG expression was observed following incubation of reporter cells with TREX1 D18N T cells for 24 hours, indicating that a soluble factor produced by these cells can induce modest ISG expression (Figure 5B). Induction of reporter ISGs was augmented by pre-incubation of TREX1 WT or D18N T cells with DMXAA, again confirming the functionality of the STING pathway in T cells. Interestingly, DMXAA-stimulated TREX1 D18N T cells produced a 5-fold more potent reporter ISG response than WT T cells (Figure 5B), indicating more robust IFN-I production and consistent with our pTBK1/IRF3 data (Figure 4C–F). To confirm that these effects were dependent on IFN-I sensing, we pre-incubated reporter cells with α-IFNAR antibody before measuring reporter ISG expression. IFNAR blockade returned reporter ISG expression to near baseline using both unstimulated and DMXAA-stimulated source cells, indicating that T cell-produced IFN-I was the source of ISG induction in both conditions (Figure 5B). Together, these results demonstrate that small molecule activation of STING or sensing of unprocessed TREX1 nucleic acid substrates can lead to IFN-I protein synthesis and secretion in T cells.

Figure 5. Activated TREX1 D18N T cells produce IFN-α protein.

(A) Setup for T cell transwell experiments (UT: Untreated, DX: Source cells + 10 μg/mL DMXAA, αIF: reporter cells + 20 μg/mL IFNAR blockading antibody, DX-αIF: source cells + DMXAA and reporter cells + IFNAR blockading antibody). Treated cells were kept in a separate plate, washed extensively after incubation, then transferred to the transwell. (B) Averaged fold induction of IFIT1, IFIT3, and USP18 in reporter WT T cells after 24 hour incubation with WT or D18N “source” T cells (3 independent experiments, 3–5 mice of each genotype, normalized to UT WT source). (C) Averaged fold induction of IFIT1, IFIT3, ISG15, IRF7, and OAS1a in bulk WT and D18N T cells cultured for 72 hours, either without or with (D) stimulation with 2e5 α-CD3/CD28-coated beads (αIF: 20 μg/mL IFNAR-blockading antibody, ISO: isotype control for α-IFNAR). (E, F) ISG expression over time in sorted naïve (CD62L⁺ CD44^low) CD4 or CD8 T cells, stimulated with 5 μg/mL plate-bound α-CD3/CD28. For (C-F), all measurements are normalized to WT cells at each time point. (G) ELISA measurement of IFN-α secreted by WT and D18N T cells after 24 hours in vitro. (H) IFN-α production of sorted WT and D18N pDCs after 24 hour culture. Panels (C-H) represent 2–3 independent experiments with 3–6 mixed sex mice 8–12 weeks of age. Error bars represent SD. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001 by unpaired student’s t-test (B, G, H) or 2-way ANOVA with Dunnett’s post-hoc test (C, D). For (C-F), significance is shown only for 72 or 96 hours, but similar statistical differences were present at 24 and 48 hours.

To further confirm that TREX1 D18N T cells produce IFN-I protein, T cells were cultured for 72 hours, and ISG expression was assessed at regular intervals. We reasoned that if TREX1 D18N T cells could not degrade self-derived DNA and were producing IFN-I as a result, then the ISG signature would continue in vitro, and would be sensitive to IFNAR blockade. We noted that ISG expression in unstimulated TREX1 D18N T cells fell relative to WT cells over the initial 24 hours, but then stabilized at a 5-fold relative induction over subsequent days. Surprisingly, this diminished ISG signature was not sensitive to IFNAR blockade (Figure 5C), suggesting that it is not dependent on IFN-I. This may indicate that resting TREX1 D18N T cells require additional signals to stimulate IFN-I production or ISG expression in vitro, and that unstimulated ISG induction measured in the transwell system (Figure 5B) represented the residual effect of deteriorating IFN-I production. A recent report demonstrated that TCR ligation can amplify STING signaling in T cells (51). As such, we examined ISG expression in T cell cultures activated with α-CD3/CD28 over the same time frame. Activated TREX1 D18N T cells maintained higher ISG expression relative to unstimulated cells, and this signature was significantly diminished by IFNAR blockade (Figure 5D). To determine if this evidence of IFN-I production was dependent on a specific T cell subset or the presence of differentiated cells, we sorted naïve CD4 and CD8 T cells to >95% purity, and repeated the experiment. ISG expression in both populations remained elevated and sensitive to IFNAR blockade (Figure 5E–F), indicating that activation-dependent IFN-I production does not require already differentiated cells, and occurs in both CD4 and CD8 T cells. Interestingly, IFNAR blockade in activated TREX1 D18N cultures did not return ISGs to WT levels, but lowered expression to levels similar to unstimulated cultures, suggesting that a component of the ISG signature is IFNAR-independent in both conditions.

To confirm that IFNAR-dependent ISG expression in activated cultures truly reflected IFN-I protein production, we assayed T cell culture supernatants for IFN-α. Unsurprisingly, IFN-α was below the limit of detection in unstimulated and CpG DNA-stimulated wells. Consistent with our pTBK1/IRF3 and transwell data, DMXAA-stimulated T cells produced measurable quantities of IFN-α, and TREX1 D18N T cells secreted more protein than WT. Importantly, activated TREX1 D18N T cells, but not WT, produced detectable quantities of IFN-α protein (Figure 5G). We compared the magnitude of these responses to TREX1 WT and D18N pDCs, as the canonical primary IFN-α producing cells. pDCs stimulated with DMXAA or CpG DNA produced substantially more IFN-α on a per-cell basis than T cells, as expected (Figure 5H). Interestingly, WT pDCs produced significantly more IFN-α in response to CpG than to DMXAA, but this response was inverted in D18N pDCs, potentially indicating alteration of STING and TLR signaling pathways due to chronic inflammatory signaling. Importantly, however, we observed no measurable spontaneous IFN-α production in TREX1 D18N pDCs. Thus, activated TREX1 D18N T cells likely represent a major source of IFN-α in the TREX1 D18N mouse.

Discussion

Failed processing of self-DNA triggers cytosolic DNA sensors, inflammation, and autoimmunity. Mice expressing catalytically inactive TREX1 D18N fail to process cytosolic DNA, resulting in inflammation and spontaneous lupus-like disease (20). Here, we show that TREX1 inactivity in T cells causes IFN-I signaling and production. TREX1 D18N mice exhibit IFNAR-dependent ISG signaling and expansion of differentiated lymphocytes. The cellular correlates of autoimmunity require TREX1 inactivity only in hematopoietic cells, indicating that TREX1-mediated DNA disposal must occur within immune cells to prevent autoinflammation. Unprocessed DNA in TREX1 D18N T cells is sensed through cGAS-STING, and activated T cells aberrantly produce IFN-I protein. Thus, TREX1 DNA degradation in T cells is critical to prevent cytosolic self-DNA sensing and spontaneous IFN-I production, providing important new insights into DNA processing and innate immunity in T cells.

The TREX1 D18N mice reveal that catalytic inactivity within bone marrow-derived cells is the primary determinant of autoinflammation. TREX1 is expressed in all murine tissues tested, suggesting that TREX1-mediated DNA disposal occurs in many cell populations (20, 52). Indeed, our data indicate that hematopoietic or somatic TREX1 inactivation induces modest ISG expression (Figure 2B), potentially reflecting IFN-I production by many cell populations following sensing of damaged DNA (13), retroelements (14), or unprocessed erythroblast DNA (53). It is therefore curious that only hematopoietic TREX1 inactivation results in indications of autoimmunity. This may reflect hematopoietic IFN-α production as the key pathogenic isoform driving autoimmunity in TREX1 D18N mice. Supporting this concept, fibroblast-specific TREX1 deletion and global SAMHD1 deletion in mice cause IFN-β-dependent ISG expression, but no evidence of autoimmunity (54–56), suggesting that IFN-β is not sufficient to cause interferonopathy. In contrast, IFN-α appears to be the primary pathogenic isoform in the NZB model of lupus-like disease (57), demonstrating its potential as a pro-autoimmune signal. We propose that somatic TREX1 inactivation induces IFN-β, whereas hematopoietic TREX1 inactivation induces IFN-α and -β synthesis to promote autoimmunity. Alternatively, it may be that IFN-I is necessary but not sufficient to promote loss of self-tolerance. This is supported by the fact that an adenoviral vector encoding IFN-α was sufficient to induce autoantibodies in NZBxW/F1 mice, but not WT mice (58). TREX1 inactivity and chronic DNA sensing in immune cells may stimulate production of other STING-inducible cytokines, such as TNF-α (59) or IL-6 (60), which could work in tandem with IFN-I to potentiate loss of self-tolerance.

TREX1 D18N immune cells express IFN-α and -β. Surprisingly, T cells exhibit more robust IFN-I expression than innate immune cells, and we observe no evidence of a spontaneous pDC IFN-I response (Figures 3–5). This finding challenges conventional thought that pDCs are the predominant source of IFN-I signaling in lupus-like disease (61, 62), and might more accurately reflect the large capacity for IFN-α synthesis by pDCs during viral infection and through TLR-dependent sensing mechanisms (46). Indeed, CpG-stimulated WT pDCs produced substantial quantities of IFN-α (Figure 5H). In contrast, TREX1 D18N pDCs produced significantly more IFN-α in response to DMXAA than to CpG (Figure 5H), despite exhibiting weaker phosphorylation of STING signaling intermediates following DMXAA treatment (Figure 4C–F). Chronic inflammation in the TREX1 D18N mouse may alter the regulation of pDC DNA-sensing pathways, augmenting some but dampening others, which may suggest that pDC-dependent responses against DNA viruses are altered in the context of interferonopathy. While this finding merits further study, we observed no evidence of spontaneous STING activation or IFN-α production in TREX1 D18N pDCs, suggesting that they likely do not measurably contribute to systemic IFN-I signaling in this model system. Other TREX1 D18N innate immune populations such as macrophages or conventional DCs likely do sense self-DNA and produce IFN-α/β (Figure 3D), but T cells consistently exhibited the most robust signature of IFN-I expression and are highly abundant relative to these cell types, suggesting that they may be the predominant source in vivo.

TREX1-mediated autoimmunity requires cGAS-STING activation (27). Our finding of IFN-α/β induction in TREX1 D18N T cells therefore requires that these cells possess a functioning cGAS-STING axis. We have shown that T cells express all molecular components of the cGAS-STING pathway and that chemical activation of STING induces TBK1/IRF3 phosphorylation (Figure 4) and IFN-α production (Figure 5G), supporting the functionality of this innate immune pathway in T cells (49–51). Further, we demonstrate that TREX1 D18N T cells are actively sensing DNA through cGAS-STING, as indicated by increased basal TBK1/IRF3 phosphorylation (Figure 4C–F). Thus, failure to process DNA in TREX1 D18N T cells triggers cGAS to activate fully functional STING, leading to IFN-I production.

Chronic IFN-I signaling in TREX1 D18N mice likely sustains high-level IRF7 expression in T cells, licensing them to produce IFN-α following DNA sensing. TREX1 D18N T cells produce considerably more IFN-I than WT cells upon STING activation (Figure 5G). This likely results from chronic overexpression of the IFN-α regulator IRF7 (Figure 4A–B). Innate immune cells require cGAS-STING to produce IFN-α in response to cytosolic DNA (35, 60), supporting the idea that STING activation results in IRF7 phosphorylation and activation. During homeostasis, IRF7 is expressed at high levels only in innate immune cells, priming them to produce IFN-α in response to nucleic acid sensing. IRF7 is also an ISG, however, creating a positive feedback loop of IFN-α production (47). Self-DNA sensing in TREX1 D18N T cells may therefore perpetually maintain this feedback loop, allowing chronic IFN-α production.

IFN-α production by TREX1-deficient T cells likely contributes to autoimmunity. IFN-I is a cytokine “signal 3” in the 3-signal hypothesis of T cell activation, promoting survival, expansion, and function of activated cells (33, 63–65). Inappropriate IFN-α production by TREX1 D18N T cells following TCR stimulation (Figure 5G) may therefore have important consequences for autoreactive T cell responses. IFN-I production in response to the STING agonist cGAMP is strongly augmented by TCR ligation in WT T cells (51). Our data indicate a similar relationship between cGAS-STING-dependent sensing of undegraded TREX1 substrates and TCR signaling. We would propose that TCR engagement in vivo stimulates TREX1 D18N T cells to aberrantly produce IFN-α/β, which may act as a self-provided inflammatory signal to facilitate expansion and survival of autoreactive clones. Potentially consistent with this model, TREX1^−/− mice lacking T cells never develop autoimmune disease, while those lacking B cells do (27), and deletion of RAG2 reduces IFN-β expression in TREX1^−/− hearts (14). Curiously, we noted that TREX1 D18N naïve T cells exhibited the highest transcriptional signature of IFN-α ex vivo (Figure 3F), but the ISG signature decayed in unstimulated cells in vitro (Figure 5C). TCR ligation was sufficient to induce in vitro IFN-I production in naïve T cells, however (Figure 5E–F), suggesting that TCR contacts may indeed be critical for driving naïve T cell IFN-α expression in vivo. Interestingly, we also noted that in vitro ISG expression in unstimulated TREX1 D18N cultures decayed but never reached WT levels, and was resistant to IFNAR blockade. Similarly, ISG expression in TCR-stimulated cultures was only partially IFNAR-dependent (Figure 5C–F). This could suggest that in addition to IFN-I, TREX1 D18N T cells produce and sense IFN-III, which signals through an alternate receptor and can also drive ISG expression (66, 67).

In summary, our findings demonstrate that T cells act as an important source of IFN-α in a model of IFN-I dependent lupus-like autoimmune disease. T cells exhibited a signature of IFN-α/β overexpression, evidence of spontaneous TBK1/IRF3 phosphorylation, and IFN-α protein synthesis following TCR stimulation in vitro. Although the mass of IFN-α protein produced on a per-cell basis is small, T cells are highly abundant relative to rare innate immune cells such as pDCs, suggesting that they possess the cumulative capacity to meaningfully contribute to systemic IFN-I levels. We propose a model in which chronic cGAS-STING-dependent DNA sensing in TREX1-deficient T cells, in conjunction with contact with self- or foreign antigen, induces inappropriate IFN-α production to potentiate autoimmunity. These findings add to our understanding of DNA sensing and innate immunity in T cells, and may have relevance to the pathogenesis of human interferonopathy.

Key points.

• TREX1 catalytic inactivity causes self-DNA sensing and IFN-I-dependent autoimmunity

• TREX1-deficient T cells exhibit evidence of DNA-sensing and IFN-α expression

• Activated TREX1-deficient T cells spontaneously produce IFN-α protein

Abbreviations

AGS

Aicardi-Goutieres Syndrome

cGAS

Cyclic GMP-AMP Synthase

DC

Dendritic cell

DMXAA

5,6-dimethylxanthenone-4-acetic acid

E/M

Effector/memory

FCL

Familial Chilblain Lupus

IFN-I

Type I interferon

IFNAR

Interferon-α/β receptor

IRF

Interferon-regulatory factor

ISG

Interferon-stimulated gene

LCMV

Lymphocytic choriomenigitis virus

MEFs

Murine embryonic fibroblasts

pDC

Plasmacytoid dendritic cell

PRR

Pattern recognition receptor

RVCL

Retinal vasculopathy with cerebral leukodystrophy

STING

Stimulator of Interferon Genes

SLE

Systemic lupus erythematosus

TCR

T cell receptor

TREX1

Three-prime Repair Exonuclease 1