ADAR1 haploinsufficiency and sustained picornaviral RdRp dsRNA synthesis synergize to dysregulate RNA editing and cause multi-system interferonopathy

ABSTRACT

Sensing of viral double-stranded RNA (dsRNA) by MDA5 triggers abundant but transient interferon-stimulated gene (ISGs) expression. If dsRNA synthesis is made persistent by transgenically expressing a picornaviral RNA-dependent RNA polymerase (RdRp) in mice, lifelong MDA5-MAVS pathway activation and marked, global ISG upregulation result. This confers robust protection from viral diseases, but in contrast to numerous other chronic MDA5 hyperactivation states, the mice suffer no autoimmune or other health consequences. Here, we find that they further confound expectations by being resistant to a strong autoimmunity (lupus) provocation. However, knockout of one allele of Adar breaks the autoinflammation-protected state of RdRp^tg mice and results in a severe disease that resembles interferonopathies caused by MDA5 gain-of-function protein mutations. Adar^+/– mice are healthy, but Adar^+/– RdRp^tg mice have shortened lifespan, stunted growth, premature fur graying, poorly developed teeth, skeletal abnormalities, and extreme ISG elevations. A-to-I edits are both abnormally distributed and increased (numbers of genes and sites). These results, with a nucleic acid-triggered and MDA5-wild-type model, illuminate the ADAR1-MDA5 axis in the regulation of innate immunity and establish that viral polymerase-sourced dsRNA can drive autoinflammatory disease pathogenesis.

IMPORTANCE

RNA virus double-stranded RNAs (dsRNAs) are important pathogen-associated molecular patterns that are sensed by the RIG-I-like receptor MDA5, which triggers an acute innate immune response involving many interferon-stimulated genes (ISGs). One key to a healthy innate immune system is that MDA5 does not sense endogenous dsRNA. This is normally ensured by dsRNA duplex-disrupting ADAR1 editing of host dsRNAs. Picornavirus RdRp^tg mice have an unusual constitutive MDA5 activation state, with very high lifelong MDA5-mediated ISG expression that confers robust protection from diverse lethal viruses. Importantly, and in contrast to numerous other chronic MDA5 hyperactivation states, the mice develop no autoinflammatory consequences. If we delete one ADAR1 allele, however, which by itself is well tolerated, the mice develop a multisystem disease that resembles the human interferonopathy Singleton-Merten syndrome. In contrast to other MDA5/ADAR1 disease models, the MDA5 and ADAR1 proteins are both wild type in this dsRNA-driven model.

INTRODUCTION

Double-stranded RNA (dsRNA), a requisite intermediate during the replication of RNA viruses, is a pathogen-associated molecular pattern that induces an array of first-line antiviral defenses. Two specialized RNA helicases function as the main pattern recognition receptors (PRRs) that detect viral dsRNA: retinoic acid inducible gene-I (RIG-I) and melanoma differentiation antigen 5 (MDA5). Both of these RIG-I-like receptors (RLRs) are cytosolic sensors that signal through the adaptor MAVS, which instigates signaling cascades that eventuate in expression of numerous interferon-stimulated genes (ISGs), either directly or via induction of secreted type I interferons (IFN-I) (1). RIG-I in general detects shorter or 5’-ppp-containing dsRNAs, and MDA5 detects the internal segments of longer (>1 to 2 kb) dsRNAs (2–4). A few RNA viruses are exclusively sensed by one RLR (1). Picornaviral dsRNAs, for example, are detected by MDA5 and not RIG-I (1, 5–8).

The challenge such a system poses for the host is to securely differentiate viral dsRNAs from the vast, diverse pool of cellular RNAs, many of which can also harbor extended RNA duplex segments, particularly within retroelement-derived RNAs that are abundant in mammalian genomes (9, 10). Inappropriate or sustained activations of RLRs and other signaling pathways by endogenous nucleic acid ligands are associated with a variety of autoimmune diseases, which include classical, common syndromes such as systemic lupus erythematosus (SLE), type I diabetes mellitus, and psoriasis as well as various and frequently severe genetic conditions known collectively as interferonopathies. These autoinflammatory conditions are characterized by persistent IFN-I and ISG expression and diverse end organ pathologies (5, 11, 12).

The RNA-modifying enzyme Adenosine Deaminase Acting on RNA 1 (ADAR1) has recently been identified as a central constraint on dsRNA-triggered autoimmunity (13–17). Encoded by ADAR, ADAR1 catalyzes post-transcriptional editing of host RNAs by deaminating adenosine to inosine, thereby disrupting A-U base-pairing and preventing formation of long uninterrupted RNA duplexes (9, 10, 18, 19). A-to-I editing most prominently prevents detection by MDA5 or protein kinase regulated by dsRNA of endogenous retroelement transcripts and other dsRNAs, which averts inappropriate immune activation or translational shutoff (9, 10). Such editing of transcripts can also cause protein recoding (17, 20, 21), although the vast majority of A-to-I edits in mammals are in non-coding regions (22–25). Strikingly, physiologically essential A-to-I editing represents a very small fraction of the editome, and moreover, most editing is unnecessary for murine homeostasis in the absence of MDA5 (26, 27). Hereditary mutations in ADAR, as well as in other genes such as TREX1, SAMHD1, RNASEH2, RNU7, LSM11, and IFIH1 (MDA5), have been linked to the development of the rare congenital inflammatory disorder Aicardi-Goutières syndrome (AGS), which clinically mimics encephalopathies caused by in utero-acquired virus infections (11, 14, 28–33). Homozygous Adar gene knockout is embryonic lethal in mice, causing mass apoptosis of fetal liver hematopoietic cells by embryonic day 11.5–12.5 (20, 34). In contrast, Adar^+/– mice are phenotypically normal and born at expected Mendelian ratios (18). Early fetal demise of Adar^−/− mice has been linked mechanistically to activation of the MDA5 pathway by dsRNA regions in cellular RNAs, chiefly repetitive elements, that are normally masked by A-I editing (15, 18, 20, 34, 35).

Indeed, constitutive MDA5 activation has been repeatedly observed to cause autoimmune syndromes such as AGS in both mice and humans (31, 36). In distinctive counterpoint, we have shown in a mouse model that chronic, systemic MDA5 activation caused by viral polymerase-generated dsRNA, which causes marked, lifelong ISG upregulation, can be well-tolerated, even when it is also strongly protective against viral diseases (37–39). The mice are transgenic for the RNA-dependent RNA polymerase (RdRp) of a neurovirulent mouse picornavirus (Theiler’s murine encephalomyelitis virus, TMEV) expressed under transcriptional control of the non-selective ubiquitin C promoter, which causes chronic dsRNA-triggered innate immune activation (37, 38). Tissues of the RdRp transgenic mice (RdRp^tg mice; see first Materials and Methods section for nomenclature) express low levels of the polymerase, which templates on host RNAs to synthesize dsRNA. Elevated dsRNA is detected in RdRp^tg mouse tissues using the K1 anti-dsRNA antibody (37). In addition, a catalytic center mutant of the TMEV RdRp lacked all ISG up-regulating activity (37). RdRp^tg mice have global, high upregulations of ISGs and robust protection against ordinarily lethal challenges by a variety of RNA and DNA viruses, including Theiler’s virus itself, encephalomyocarditis virus (EMCV), vesicular stomatitis virus, a DNA herpesvirus (pseudorabies virus), and Friend retrovirus (37, 38, 40, 41). The model is mouse strain independent, with equivalent phenotypes in FVB/NJ, BALB/c, and C57BL/6J mice.

RdRp^tg/– and RdRp^tg/tg mice develop normally, with the onset of the major ISG expression profile shortly after birth (39). The intriguing lack of deleterious effects from their chronically elevated ISGome differs strikingly from other constitutive MDA5 activation states (36). Body size, morphology, organ histology, and longevity are normal (37–39). Crosses with Ifih1^−/− mice showed that the sustained innate immune activation is strictly dependent on MDA5, which is congruent with MDA5 but not RIG-I being the sensor of picornaviral dsRNAs (1, 5–8). It further depends on the downstream adaptor MAVS and the type I IFN receptor (IFNAR1) and is abolished by knockout of these genes (37). Although type I IFNs are not detectably overexpressed in RdRp^tg mice tissues, antibody-mediated blockade of the type I IFN receptor in adults terminates the ISG profile, indicating that some ongoing IFN-I signaling is required to sustain it (38, 39). TLR3, IFNGR1, and RAG1 knockout crosses also showed that the ISG profile does not depend on TLR3, interferon gamma signaling, or the adaptive immune system. As expected from the complete abrogation of ISG upregulation by Ifih1 knockout, Ddx58 knockout mouse crosses further confirmed no dependence on RIG-I (our unpublished data). Thus, the model is distinctive in representing a pure dsRNA-induced MDA5 hyperactivation state, mediated through the wild-type (WT) MDA5 sensor and “viral” dsRNA.

Prompted by the well-tolerated ISG elevations in this model, we here carried out experiments that demonstrate that RdRp^tg mice also resist the induction of SLE in the BM12 lupus model, which we find is linked to increased quantities and effector function of regulatory T cells (T_reg). We show that introducing a single Adar allele knockout—which, similar to RdRp transgenesis, produces no abnormalities by itself—breaks the RdRp^tg protective state, yielding a severe autoinflammatory disease. RdRp^tg mice lacking one Adar allele have stunted growth, gray fur, abnormal dental and skeletal structures, failure to thrive, highly dysregulated ISG expression, and abnormal A-to-I editing. This dsRNA-driven, MDA5-wild-type model establishes that viral polymerase-sourced dsRNA can drive interferonopathy pathogenesis and illuminates the autoimmunity-preventing role of ADAR1.

RESULTS

RdRp^tg mice resist the induction of SLE

Since picornaviral RdRp^tg mice do not exhibit apparent fitness “costs” of their chronic immune system activation (37–39), an unanswered question is whether they may be more autoimmunity-prone like other constitutive MDA5 activation models (36, 42, 43). To address this experimentally, we carried out a strong autoimmunity provocation by adopting the BM12 inducible SLE model, in which lupus is initiated by injection of splenocytes from MHC class II-mismatched mice (44, 45). WT and RdRp^tg/– mice were injected with 1 × 10⁸ splenocytes isolated from BM12 mice or were control injected with phosphate buffered saline (PBS) and evaluated for disease 2 weeks later. One of the clearest morphological changes in the BM12 model is splenomegaly, which results from germinal center expansion. Spleens were harvested and weighed before processing for flow cytometric analyses. As anticipated, we observed a significant increase in spleen size in WT mouse BM12 splenocyte recipients, indicative of lupus-like disease initiation (Fig. 1A). In contrast, RdRp^tg mouse BM12 splenocyte recipients had only minor, statistically non-significant increases in spleen size (Fig. 1A; P = 0.126 as compared with PBS-injected RdRp^tg mice).

Fig 1.

RdRp^tg mice resist SLE-like disease induction in the BM12 model of lupus. Ten-week-old mice were challenged with 100 million BM12-derived splenocytes. Mice were harvested 14 days post-challenge for analyses. (A) Splenomegaly. The weights of BM12 and sham-injected mice’s spleens were measured. (B and C) Germinal center B cells, plasma cells, and Tfh cells. RBC-lysed, single-cell suspensions of splenocytes from BM12 or sham-injected animals were generated for flow staining and germinal center B cells and plasma B cells (B), or follicular helper cell populations (C) were measured by flow cytometry according to the gating described in Materials and Methods. (D) Anti-nuclear antibodies. Sera from BM12 splenocyte-injected or sham-injected mice were used to measure anti-double-stranded DNA (dsDNA) antibodies and anti-SmAg antibodies by quantitative ELISA. Data in panels A–D are from n = 7, 6, 8, and 8 mice for WT (PBS), RdRp^tg/– (PBS), WT (BM12), and RdRp^tg/– (BM12) groups, respectively. (E) T_reg cells. RBC-lysed single-cell suspensions were generated from spleens of untreated, 10-week-old WT or RdRp^tg mice (n = 5 for each group), and T_reg subsets were determined by flow cytometry. Data were analyzed using one-way analysis of variance (ANOVA), followed by Tukey tests for (A–D) and an unpaired Student’s T test for (E). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Data points represent individual animals, and graphs show means with SD.

We next examined cellular subsets in the spleen, specifically germinal center B cells and plasma B cells, which expand as they become activated to secrete anti-nuclear antibodies (ANAs), and T follicular helper (Tfh_fh) cells, which provide antigenic stimulus in germinal center reactions (46, 47). For all three cell populations, WT mouse BM12 splenocyte recipients developed significant increases, indicative of ongoing germinal center reactions and increased antibody production (Fig. 1B and C). In contrast, and paralleling the spleen measurements, RdRp^tg mouse BM12 splenocyte recipients developed little to no increases in all three cellular populations. Significant increases over sham-treated animals were only seen in germinal center B cells (Fig. 1B and C). Minor changes in overall B cell and T cell populations were also observed, but none that account for the drastic differences in cellular subsets seen in WT BM12 splenocyte recipient animals (Fig. S1A and B).

To corroborate the results, we collected serum from splenocyte recipients and control PBS recipients to measure production of ANAs, which are a hallmark of SLE and drive pathology in this model (44, 45). We measured ANAs against double-stranded DNA (dsDNA) and Smith antigen (SmAg), both of which are frequently observed in SLE patients and SLE mouse models (47, 48) (Fig. 1D). Similar to the changes in cellular populations, there were significant increases in both of these autoantibodies in WT BM12 splenocyte recipients compared to the PBS recipient controls. In contrast, the RdRp^tg mouse BM12 splenocyte recipients produced less of both autoantibodies.

T_reg cell expansion

Our previous investigations did not reveal significant differences in immune cell subsets between WT and RdRp^tg mice (38). Here, we extended the studies to assess autoimmune suppressor cells, specifically T_regs, which are critical for maintaining immune-tissue homeostasis. Knockout of T_regs results in disseminated autoimmune disease followed by rapid organismal decline and death (49, 50). We measured T_reg cell subsets in age-matched animals. There was a highly significant increase in T_reg cells in RdRp^tg/– mice compared to WT and specifically an increase in mature, effector T_regs (Fig. 1E). T_reg cell expansion in RdRp^tg mice may aid their tolerance to chronic innate immune activation.

The combined cellular and ANA data indicate that RdRp^tg mice are better able to control disease induction after an autoimmune provocation than WT mice. This result is distinctive compared to other MDA5-pathway-driven hyperimmune mice models, which have been shown to be equally or more susceptible to autoimmunity induction (42, 43).

Investigation of ADAR1

In addition to T_reg cells, we suspected that mechanisms of protection at the intrinsic cellular level are likely to provide key regulation. Given its known role in preventing activation of the MDA5 pathway by double-stranded segments of endogenously encoded RNAs, we hypothesized that ADAR1 may prevent adverse inflammatory sequelae in RdRp^tg mice. First, we determined whether Adar expression is affected by RdRp genotype status. Adar encodes two functional isoforms, ADAR1 p150, which edits transcripts in the cytoplasm and is IFN inducible, and constitutively expressed ADAR1 p110, which is generated by alternative splicing or by translation from the p150 mRNA by leaky internal ribosome scanning (51) and edits transcripts in the nucleus (19, 52, 53). We observed no significant difference at either the mRNA or protein levels in the expression of the p110 isoform in RdRp^tg mice as compared to WT mice (Fig. 2A and C). In contrast, there was a marked increase in expression of the IFN-inducible p150 isoform (Fig. 2B and C). These results suggested that ADAR1 p150, itself an ISG, is upregulated in RdRp mice, perhaps helping to modulate the ISG activation in these animals.

Fig 2.

ADAR p150 and ADAR p110 levels in RdRp^tg mice and effects of ADAR knockdown on RdRp-dependent ISG expression. RNA isolated from brains of 4- to 5-week-old WT or RdRp^tg/– mice (n = 5 for WT, 6 for RdRp^tg/–) was used to measure (A) ADAR p110 and (B) ADAR p150 transcripts by quantitative PCR (qPCR). Data points represent individual animals, and graphs show means and SD. (C) Immunoblotting for ADAR p110 and ADAR p150 in 4- to 5-week-old mouse WT and RdRp^tg/– brain. n = 3 animals per genotype. (D) A549 cells with inducible Theiler’s virus RdRp (Tet-on system). Data are shown from two representative knockdowns. (E) RNAs isolated from the parallel knockdowns done in panel D were used for qPCR analysis to determine relative levels of the ISGs OAS and ISG15 mRNAs. mRNAs were harvested 54 hours after ADAR-targeting small interfering RNA (siRNA) addition and 48 hours after doxycycline (dox) addition. Comparisons were made between dox-treated and dox-untreated cells and expressed as fold mRNA changes induced by dox. Transfection controls (cells receiving transfection reagents but no siRNA) showed similar levels of ISG induction after dox treatment as control siRNA-transfected cells, indicating a lack of contribution of transfected siRNAs to immune activation. Data are means and SD of triplicate biological replicates with three technical replicates each. NS: not significant; *P < 0.05, **P < 0.01, and ****P < 0.0001; unpaired Student’s T test.

We next knocked down ADAR expression in human lung epithelial cells (A549) in which we have engineered and validated inducible TMEV RdRp expression under the control of a doxycycline (dox) inducible promoter (37). These cells broadly upregulate ISGs in response to RdRp induction in a pattern very similar to the RdRp^tg mouse (37). In two separate experiments, we achieved knockdowns of ADAR p110 and p150 isoforms (Fig. 2D). Dox induction of the RdRp after siRNA transfection led to over three log₁₀ increases in mRNAs for two classical ISGs, OAS and ISG15, which were moderately accentuated by ADAR depletion in these short-term experiments (Fig. 2E). We therefore proceeded to generate mice with Adar gene knockouts.

Adar haploinsufficiency, by itself well tolerated, causes multi-system disease in RdRp^tg mice

As noted, by 2 weeks after birth, RdRp^tg mice develop high, sustained upregulations of many ISGs and subsequently tolerate them throughout their (normal) lifespans (37, 39). To determine whether ADAR1 is involved in the protection against deleterious effects of autoinflammation, RdRp^tg/tg mice were crossed with Adar^+/– mice. RdRp^tg/– mice also have elevated ISGs, albeit with more variability in fold induction than RdRp^tg/tg mice (for example, see Fig. 2B and C for expression of the ISG ADAR1 p150, and Painter et al. [37]). Since in our hands and others Adar^+/– mice have no phenotypic abnormalities, we did not anticipate a major effect in first-generation crosses even in the presence of an RdRp-mediated ISG response, as the F1 animals will retain a functional Adar allele.

However, that was not the outcome. First, there were major gross phenotypic differences between RdRp^tg/– Adar^+/– mice and littermate RdRp^tg/– Adar^+/+ controls (Fig. 3A through C). The RdRp^tg mice lacking one Adar allele were significantly smaller in size (Fig. 3A and B), with gray instead of black fur (Fig. 3A and C). These features were equivalent in males and females (Fig. 3A and C). RdRp^tg/– Adar^+/– mice also have small, poorly developed, misshapen teeth (Fig. 3C). Runting was unrelated to the dental abnormalities, since suckling behavior was unaffected. In addition, the size evaluations were done at the time of weaning, and post-weaning mice were kept on a wet chow diet, which they were also observed to ingest equivalently. Adar^+/– mice were normal in appearance as expected.

Fig 3.

Phenotypic and histological differences in RdRp^tg/– Adar^+/– mice. (A) Size and coat color differences in 5-week-old, littermate RdRp^tg/–, and RdRp^tg/– Adar^+/– mice. (B) Weight differences between 4- and 5-week-old littermate RdRp^tg/– and RdRp^tg/– Adar^+/– mice (n = 9 and 5, respectively). (C) Dental developmental differences between RdRp^tg/– and RdRp^tg/– Adar^+/– mice. (D) Survival of animals from each group, followed for 20 weeks, and shown as Kaplan-Meier plot. (E) hematoxylin and eosin (H&E)-stained tissue sections from kidney, brain, lung, liver, heart, and spleens from RdRp^tg/– and RdRp^tg/– Adar^+/– littermate mice. Pathological grading revealed no significant scoring in RdRp^tg/– Adar^+/– animals. (F) Blood urea nitrogen from serum from 4- to 5-week-old animals in WT, Adar^+/–, RdRp^tg/–, and Adar^+/– RdRp^tg/– mice. n = 5, 6, 5, and 8 for WT, Adar^+/–, RdRp^tg/–, and Adar^+/– RdRp^tg/–, respectively. (G) Anti-SmAG antibodies from serum from 4- to 5-week-old animals as measured by quantitative ELISA. Four outlier high values were seen, but two were in the WT group. n = 15, 7, 10, and 9 for WT, Adar^+/–, RdRp^tg/–, and Adar^+/– RdRp^tg/–, respectively. Anti-dsDNA antibody levels from matched serum were below the limit of detection for all animals tested. Where not indicated, all data and tissue sections come from a mix of male and female mice. Data in panel B were analyzed by two-way ANOVA, where ****P < 0.0001. Data in panel D were analyzed using a log-rank test (Mantel-Cox), where ****P < 0.0001. Data in panels F and G were analyzed by one-way ANOVA, where *P < 0.05; data points represent individual animals, and graphs show means and SD.

Because of these phenotypic differences, we carried out tissue dissections. Survival of the RdRp^tg/– Adar^+/– mice began to fall significantly by 10 weeks after birth (Fig. 3D), so these studies were done at 4–5 weeks. Soft viscera, including heart, lungs, liver, spleen, intestines, kidney, and brain, were proportionately small, but the organ morphologies were otherwise not different from WT mice. Histopathology analysis of tissue sections by a veterinary pathologist did not reveal evidence for tissue inflammation or other abnormalities (Fig. 3E; see Supplemental Methods for scoring and data). Since some MDA5-driven mouse autoimmune models manifest with kidney impairment, blood urea nitrogen (BUN) values were determined at 4–5 weeks of age. Mean BUN was slightly higher in RdRp^tg mice but was not in either Adar^+/– group (Fig. 3F). We also measured serum ANAs and found no significant increases in anti-dsDNA antibodies (data not shown as levels for all animals were below the limit of assay detection) or anti-SmAg antibodies (Fig. 3G). In addition to the abnormal teeth, necropsy revealed thin, pale long bones, suggesting decreased bone density. Bone frailty was also readily apparent during cervical dislocations, which subjectively required much less force. Similar dental and bone abnormalities can occur in Singleton-Merten syndrome (SMS), a rare human autosomal dominant innate immune disorder consisting of musculoskeletal abnormalities (osteopenia, osteoporosis, skull thickening, small stature, and ligament frailty), dental anomalies (poor, primarily anterior teeth formation), variable arterial calcification, inflammatory skin changes (psoriasis), and a characteristic facies (high anterior hairline, broadened forehead, asymmetric ptosis, smooth philtrum, and thin upper vermilion) (54, 55). Genetic studies of individuals with SMS have revealed causation by a subset of single amino acid missense gain-of-function (GOF) mutations in MDA5 (R822Q, A489T, and T331I/R), which cause constitutive activation (56–60). There is inter-familial and intra-familial variation in syndromic manifestations, with partial penetrance evident in some kindreds (55). Overlap with Aicardi-Goutières syndrome has been reported (58, 59, 61). Similarly, two families with atypical SMS lacking the dental abnormalities were found to have constitutively active single amino acid mutants of RIG-I rather than MDA5 (62).

As RdRp^tg/– Adar^+/– mice had micrognathia as well as small, misshapen, and partially translucent incisors (Fig. 3C), micro-computed tomography (µCT) imaging and mechanical property assessments of femurs were performed to determine whether RdRp^tg/– Adar^+/– mice also had altered underlying changes to bone architecture or properties. Consistent with the runting observed in 5-week-old mice (Fig. 3A and B), femurs of 6-week-old RdRp^tg/– Adar^+/– were significantly shorter than control groups (Fig. 4A and B). As in other RLR-related genetic disorders (55, 62), weight, femur length, and other mouse characteristics were variably penetrant. Bartlett’s test demonstrated a statistically significant (P = 0.0142) difference in homoscedasticity for femur length, yet a Kolmogorov-Smirnov test did not identify significant (alpha = 0.05) deviation from normal distribution for RdRp^tg/– Adar^+/– or other groups. Thus, we did not separate RdRp^tg/– Adar^+/– mice into “penetrant” and “non-penetrant” subsets in subsequent analyses and graphics, but we labeled mice with gray fur with light blue symbols in Fig. 4. RdRp^tg/– Adar^+/– femurs had significantly reduced cortical bone volume fraction and correspondingly reduced measures of cortical thickness, perimeter, area, and porosity (Fig. 4C). In contrast, trabecular number and trabecular separation (Tb.Sp) were not different between groups (Fig. 4D and E). Mechanical properties of the femurs were further investigated using three-point bending to failure. Stiffness and maximum load of RdRp^tg/– Adar^+/– femurs were significantly reduced compared to control groups (Fig. 4F and G). Modulus and ultimate stress of the femurs were calculated from mechanical properties and mid-diaphysis cross-sectional geometry from µCT. In contrast to the reduced mechanical properties observed, these material properties of RdRp^tg/– Adar^+/– femurs were not altered (Fig. 4I and J), nor was the total cortical mineral density calculated from the µCT imaging (Fig. 4H). These results indicate that double heterozygosity (RdRp^tg/– Adar^+/–) led to femurs having inferior stiffness and strength stemming from smaller size but not from impaired bone material quality, in addition to their dental maldevelopment and fur graying.

Fig 4.

Skeletal and dental features of WT, RdRp^tg/–, Adar^+/–, and Adar^+/– RdRp^tg/– mice. Femurs from 6-week-old animals were manually de-fleshed and used for all analyses. (A) Femur lengths of WT, Adar^+/–, RdRp^tg/–, and RdRp^tg/– Adar^+/– mice (n = 8, 6, 10, and 17 animals, respectively). (B) Representative µCT images, distal femur sections, and whole bone. (C) Cortical bone volume fraction (BV/TV), n = 8, 7, 10, and 17. (D) Trabecular number, n = 8, 7, 9, and 17. (E) Trabecular separation, n = 8, 7, 9, and 17. (F) Stiffness, n = 8, 7, 10, and 15. (G) Maximum load, n = 8, 7, 10, and 17. (H) Cortical total mineral density, n = 8, 7, 10, and 17. (I) Modulus, n = 8, 7, 10, and 17. (J) Ultimate stress, n = 8, 7, 10, and 17. Mice with gray fur are indicated by light-blue symbols. Data were analyzed first by using a ROUT test to remove outliers (Q = 1%), which resulted in the removal of two mice from one group in one panel (the RdRp^tg/– Adar^+/– group in panel F, stiffness testing; hence, there are 15 mice as opposed to the 17 for this genotype in the other panels). A one-way ANOVA comparing RdRp^tg/– Adar^+/– to each other group was used, followed by a Tukey test where *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Data points represent individual animals, and horizontal bars indicate means.

Resemblance to classical human SMS was partial, as the mice did not develop aortic or cardiac valvular calcification—a core feature of the human syndrome—nor was there aberrant brain calcification (Fig. S2), as seen in AGS (11) and in some SMS-AGS overlap cases (58, 59, 61). Psoriasis is less commonly observed in SMS patients and was not observed in RdRp^tg/– Adar^+/– animals, leaving fur graying as the identifiable integument abnormality. Glaucoma can also occur in SMS, but intraocular pressures measured by tonometry in affected mice were normal and not significantly different from WT mice (data not shown).

Transcriptional changes in RdRp^tg/– Adar^+/– mice are consistent with an interferonopathy

The SMS-resembling phenotype triggered by the combination of one RdRp transgene allele and one Adar null allele, with neither alone producing abnormalities, prompted us to examine tissues further for differences that may be driving pathogenesis. We performed RNA-seq on brain tissue, with four biological replicates from each of the four genotypes WT, Adar^+/–, RdRp^tg/–, and RdRp^tg/– Adar^+/–. Brains were analyzed because we previously observed that brains of RdRp^tg mice have particularly high ISG elevations compared to other organs, although gene expression changes are qualitatively similar and substantial across all major organs (39). Here, in multidimensional scaling plots, we observed that the samples cluster strongly by genotype along the M1 axis, which validates that animals of the same genotype have similar gene expression profiles (Fig. 5A). The double heterozygotes (RdRp^tg/– Adar^+/–) cluster the farthest from any other genotype, suggesting that they have distinctly different RNA expression patterns. Assessing expression changes across the groups, the greatest differences were between RdRp^tg/– Adar^+/– mice relative to WT mice and relative to Adar^+/– mice, with the majority of upregulated genes being known ISGs (136 out of 151 and 124 out of 137, respectively; Fig. 5B and C, Fig. S3). These outcomes suggest that RdRp^tg/– Adar^+/– mice have an extreme ISG response that likely leads to the Singleton-Merten interferonopathy-resembling phenotype. Interestingly, when we examined which genes are changing across genotype comparisons, we saw a largely overlapping set of ISGs upregulated in the presence of the RdRp transgene or in the absence of one Adar allele (Fig. 5D). All three non-WT genotypes (Adar^+/–, RdRp^tg/–, and RdRp^tg/– Adar^+/–) showed significant upregulation of overlapping ISGs, but the magnitude of upregulation varies drastically by genotype (Fig. 5D). Consistently, the dual heterozygotes have extremely high induction, elevated three- to fivefold over what we observed in RdRp^tg/– mice, although largely concordant in which genes are induced, while Adar^+/– have only low-level ISG induction (Fig. 5B and D). Thus, synergy between the RdRp transgene and the absence of one Adar allele leads to a hyper-induced ISG state in mice that is no longer tolerated. We considered whether differing expression of the TMEV RdRp transcript itself might be influencing the observed double heterozygote differences (even though the ubiquitin C promoter controls its transcription), but the RdRp mRNA was similarly elevated in both (Fig. 5E).

Fig 5.

RNA-seq analysis reveals a clear pattern of ISG upregulation suggestive of an interferonopathy. RNA-seq from neuronal tissue from n = 4 age-matched (5 weeks) mice per group. Where possible, littermate controls were used for analysis. Two male and two female mice were used per group. (A) Multi-dimensional scaling plot of RNA expression profiles in WT-WT, Adar^+/–, RdRp^tg/–, and RdRp^tg/– Adar^+/– mice. (B) Heatmap of differentially expressed ISGs across the four genotypes (Fig. S3 shows an enlarged version with gene names annotated). FPKM values were log transformed with one pseudocount to facilitate visualization. (C) Summary table of differentially expressed genes (DEGs) and proportions of which are known ISGs as determined by the Interferome database. (D) Differentially expressed genes that are shared across all group comparisons. (E) RdRp mRNA abundance in the four genotypes. (F) Canonical molecular pathways from IPA that are significant across group comparisons.

Pathway analysis

Ingenuity pathway analysis (IPA) was used to assess which canonical molecular pathways are activated or inhibited. Adar^+/–, RdRp^tg/–, and RdRp^tg/– Adar^+/– mice showed activation of genes related to interferon signaling, neuroinflammatory signaling, and PRRs (Fig. 5F). Genes associated with dendritic cell (DC) maturation were activated in RdRp^tg/– and RdRp^tg/– Adar^+/– mice (Fig. 5F). Increased DC maturation, specifically in RdRp^tg/– Adar^+/– animals, may contribute to the observed disease. We also carried out analyses of upstream regulators that may be responsible for gene expression patterns. In WT-Adar^+/–, RdRp^tg/– Adar^+/+, and RdRp^tg/– Adar^+/– mice, we not surprisingly identified Ddx58 (RIG-I) and Ifih1 (MDA5) as upstream regulators activated in those samples (Fig. S4). Main intermediates in interferon signaling pathways, including IRF7, IRF9, STAT1, and STAT2, were also statistically significant predicted key regulators (Fig. S4). Predicted regulation by additional PRR proteins was identified only in mice missing one Adar allele (Adar^+/– and RdRp^tg/– Adar^+/– mice), including TLR3, which recognizes dsRNA, and ZBP1 (DAI), which recognizes Z-form nucleic acids (63), both resulting in activation of interferon signaling (Fig. S4). Interactions with other nucleic acid-sensing molecules besides MDA5 may thus play roles in the sensing of ADAR1-edited transcripts.

Molecular and cellular differences in RdRp^tg/– Adar^+/– mice validate RNA-seq analyses and are consistent with interferonopathy

To characterize the RNA-seq results further, we determined levels of ISG mRNAs in brain tissues of Adar^+/+ and Adar^+/– mice, with or without the RdRp transgene. Similar to the RNA-seq data, in Adar^+/– mice, there were only small increases in levels of three classical ISG mRNAs compared to WT mice (about two- to fivefold; Fig. 6A). In contrast, and consistent with prior data (37–39), RdRp^tg/– mice had major increases compared to WT mice. However, the dual heterozygotes (RdRp^tg/– Adar^+/–) had much higher ISG mRNA elevations, on the order of 200- to 600-fold relative to WT.

Fig 6.

ISG and leukocyte subset differences in RdRp^tg/– Adar^+/– mice. (A) qPCR of three representative ISGs (Ifit1, Isg15, and Oasl2) in brain tissue of 4- to 5-week-old WT, Adar^+/–, RdRp^tg/–, and RdRp^tg/– Adar^+/– mice (n = 5, 5, 6, and 5, respectively). (B–E) Flow cytometry analysis of cellular populations derived from the spleens of 4- to 5-week-old mice. Single-cell, RBC-lysed solutions were prepared for use in analysis. (B) BST-2 expression on the main immune cell subsets in the spleen, including T cells (CD3⁺), B cells (CD19⁺), DCs (CD11c⁺), granulocytes (Ly6G⁺), and NK cells (NKp46⁺). n = 7, 7, 4, and 4 for WT, Adar^+/–, RdRp^tg/–, and RdRp^tg/– Adar^+/–, respectively. (C) Monocyte populations across all four groups. (D) BST-2 expression on monocyte/DC cell subsets, including monocytes, cDC1s, cDC2s, and monocyte-derived DCs (moDCs). (E) Activation (CD80/86 expression) of monocyte/DC subsets, including monocytes, cDC1s, cDC2s, and moDCs. C–E: n = 12, 7, 10, and 8 for WT, Adar^+/–, RdRp^tg/–, and RdRp^tg/– Adar^+/–, respectively. For all graphs shown, data were analyzed using a one-way ANOVA followed by a Tukey test to determine significance, where *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Data points represent individual animals with the mean and SD shown as bars.

Importantly, breeding to Ifih1^–/– mice demonstrated that loss of MDA5 resulted in rescue of animal growth and a complete abolishment of ISG upregulation observed in RdRp^tg/– Adar^+/– animals (Fig. S5), reinforcing the MDA5 dependence of the phenotypes. Given the massively amplified ISG transcriptional pattern in the double heterozygotes and the immune cell activation and maturation pathways enrichment observed (Fig. 5; Fig. S6), we next harvested the spleens of 4- to 5-week-old mice for flow cytometry analyses.

Aberrantly activated immune cell subsets are canonical autoimmune disease features, and identifying them would support the conclusion of a dysregulated immune state in RdRp^tg/– Adar^+/– mice. In major immune cell populations in the spleens, including B cells, T cells, neutrophils, DCs, and granulocytes, we found no major differences in cell population proportions across different genotypes (Fig. S7A). The antiviral protein BST-2 (Tetherin) is a classical ISG with a high dynamic range of expression in response to type I interferon stimulation in multiple mammals, including mice (64–66), and the mRNA was consistently upregulated across comparisons in the RNA-seq experiments. We, therefore, determined cell surface BST-2 protein expression in the various immune cell populations and found it to be elevated on all cell types of RdRp^tg/– Adar^+/– mice and was slightly upregulated on DCs in WT-Adar^+/– and RdRp^tg/– animals and granulocytes from RdRp^tg/– animals (Fig. 6B), similar to what we have observed previously (38). We also examined monocyte and dendritic cell differentiation, maturation, and activation in these animals (Fig. 6C, D and E). Monocytes were significantly increased in RdRp^tg/– Adar^+/– animals (Fig. 6C), and cDC1s were also slightly decreased (Fig. S7B). Monocytes are circulating precursor cells that infiltrate into the tissue upon detection of proinflammatory stimuli and differentiate into effector monocyte-derived DCs (moDCs). An increase in this cell population is suggestive of increased demand for precursor cells in these animals. In support of this hypothesis, we observed increased BST-2 expression on all DC subsets from RdRp^tg/– Adar^+/– mice, as well as increased expression of the activation markers CD80/86 on RdRp^tg/– Adar^+/– monocytes and moDCs (Fig. 6D and E). Disorders in DC function and activity have been widely implicated in several autoimmune diseases, including SLE, and may contribute to disease development in Adar^+/– RdRp^tg/– animals.

The immune cell profiling above and the upregulated ISG profiles raised the question of whether proinflammatory cytokines are elevated in the double heterozygotes. Indeed, transcripts for IFNbβ, which were undetectable at baseline in WT and Adar^+/– mice, were detectably elevated, but at low levels, in both RdRp^tg/– and more so in RdRp^tg/– Adar^+/– mice (Fig. S8). While classical proinflammatory cytokines such as IL6 and TNFα were not elevated in any group, several proinflammatory C-C and C-X-C chemokines were elevated, most prominently CXCL10, CXCL11, CCL2, and CCL5 (Fig. S8).

Editome analyses show that dual heterozygotes have dysregulated A–I editing

To interrogate the relationship of the loss of one Adar allele to the SMS-like phenotype of RdRp^tg/– Adar^+/– mice, we characterized genome-wide A-to-I editing differences between all four genotypes of animals (Fig. 7). We first carried out whole-exome sequencing to eliminate any single nucleotide polymorphisms (SNPs) that may be unique to our colony or genotypes (of which very few were detected), as compared to the reference genome. We then identified all A-to-G changes in RNA-seq data (inosine is decoded as guanine during sequencing). We found that the majority of edited sites in all four genotypes were in intergenic regions and 3′ UTRs, with a fractionally larger portion for 3′ UTRs in RdRp^tg mice. Exons and introns were represented approximately equivalently (Fig. 7A and B). Interestingly and unexpectedly, while edited site locations were similarly proportioned among the four genotypes (Fig. 7A), the number of edited sites increased substantially (50%–65%) in RdRp^tg/– Adar^+/– mice compared to the other three genotypes (Fig. 7B). Numbers of edited genes also increased. This seemingly paradoxical result in the Singleton-Merten-affected dual heterozygotes was apparent across all gene regions. In contrast, in the Adar^+/– and RdRp^tg/– mice, the number of edited genes decreased compared to WT mice, and the overall number of edited sites remained largely unchanged (Fig. 7B). Despite the number of edited sites being unchanged for WT, Adar^+/–, and RdRp^tg/– mice, there was a shift from intergenic to extragenic regions in these animals, and there was a substantial proportionate rise in 3′ UTR editing in RdRp^tg/– mice of both genotypes, which was quantitatively greater in RdRp^tg/– Adar^+/– mice.

Fig 7.

Analysis of the A-to-I editing reveals increased editing in RdRp^tg/– Adar^+/– mice. Changes in A-to-I edits were determined via comparison of RNA-seq data with whole-exome DNA sequencing. (A) Pie charts showing the proportion of editing sites within a given RNA element for each mouse genotype. (B) Table summary of the distribution of overall numbers of edited genes and sites for each genotype and the breakdown of locations of the edit sites among different RNA elements. For each of the four genotypes, four animals were sequenced, and to be counted, a gene must have been edited in all four animals sequenced, but editing can occur anywhere in the gene. For sites, identical sites must be edited in all four animals of a genotype. (C) Venn diagrams showing the overlap of edited genes (top) or sites (bottom) among the four genotypes. (D and E) Immunoblot analysis from two animals per group (D) and gene expression fold change (E) of three Adar^+/– RdRp^tg/– uniquely edited genes in the RNA-seq data set (as compared to WT) in 5-week-old neuronal tissue. Rig-I, Isg15, and Zbp1 were evaluated as they are each ISGs that are highly upregulated in mice expressing RdRp and are key regulators of the antiviral response. (F) Expression of Adar isoforms in neuronal tissue of 4- to 5-week-old mice from all four genotypes, as determined by western blot (n = 2 per group).

Because they possess only one functional Adar allele, it was unexpected that RdRp^tg/– Adar^+/– mice would have these increases in the number of A-to-I edited genes and A-to-I edited sites (Fig. 7B). Thus, there is a dysregulation of A-to-I editing. When we examined the overlap of edited sites or edited genes between genotypes (visualized in Venn diagrams; Fig. 7C), we observed the highest number of uniquely edited genes (719) and sites (695) in RdRp^tg/– Adar^+/– mice, indicating that these animals have significantly different editing patterns compared to the other genotypes. It is notable that RdRp^tg/– Adar^+/– mice have increased editing in long interspersed nuclear elements (LINEs), short interspersed nuclear elements (SINEs), DNA transposons, and other retrotransposons, since increases in retroelement transcription are associated with autoimmune disease states (67). In addition, previous studies have shown ADAR1 to be critical for the regulation of retroelement transcript self-reactivity (9, 10). Increased ADAR-specific editing of retroelements is normally protective, but in our double heterozygote mice, this alone is not compensatory enough to prevent disease, indicating that the A-to-I editing process in these animals is abnormal.

Next, we used RNA-seq to analyze the changes in edited genes, along with relative expression in each group of mice, to determine associations with activated/inhibited canonical pathways and enriched diseases and biological functions (Fig. S9). Of edited mRNAs in RdRp^tg/– Adar^+/– mice, genes associated with disease and biological functions related to apoptosis, cell death, and necrosis were enriched (Fig. S9). Notably, there was a clear reduction in genes involved in inflammatory disease states, which was unique to comparisons involving RdRp^tg/– Adar^+/– mice (Fig. S9).

To explore the question further, we asked how A-to-I editing of ISG mRNAs, when it occurs, affects expression of their encoded proteins in RdRp^tg/– Adar^+/– mice. We measured mRNA levels and performed protein immunoblotting for several ISGs that are uniquely edited in RdRp^tg/– Adar^+/– mice: Rig-I, Isg15, and Zbp1. Both Rig-I (Ddx58) and Zbp1 (Zbp1) were also predicted by the RNA-seq analysis to be key upstream gene regulators (genes with known activation/inhibition effects on the enriched pathways; Fig. S9). Expression levels of Rig-I, Isg15, and to a much lesser extent and not at the protein level, Zbp1, were increased in RdRp^tg/– and RdRp^tg/– Adar^+/– mice, but not in Adar^+/– mice (Fig. 7D and E), indicating that A-to-I editing does not reduce expression of these ISGs. Additionally, ADAR1 isoforms were expressed in RdRp^tg/– Adar^+/- mice at relatively similar levels as their wild-type counterpart (Fig. 7E and F), with slightly higher p150 protein in RdRp^tg/– mice and RdRp^tg/– Adar^+/– mice. Notably, in Adar^+/– mice alone, we observed decreased ADAR1 protein expression (Fig. 7F), as would be predicted from loss of an allele.

Endogenous retroelements are enriched in the 3′ UTRs of ISG mRNAs (68). Despite the increased ISG mRNA expression in RdRp^tg/– Adar^+/– animals compared to WT animals (Fig. 5C), the proportion of uniquely edited sites that were within an ISG RNA vs those of other genes was not significantly elevated (Fig. S10). This was the case despite the double heterozygotes having an increase in total edited genes and sites (Fig. 7C). Compared to the WT and double heterozygotes, RdRp^tg/– and particularly Adar^+/– animals had small, though significant, reductions in the proportion of uniquely edited sites falling within an ISG. These results suggest that while ISGs themselves were also edited, the amount of editing within the ISG mRNAs does not account for the observed dysregulation of Adar editing.

In summary, single Adar allelic loss coupled with transgenic expression of the picornavirus RdRp synergizes to cause an extreme ISG response, resulting in an interferonopathy remarkably similar to Singleton-Merten syndrome. This work presents a novel model for this disease, which can be further studied to unlock molecular mechanisms driving the pathology. The ADAR-intact model is exceptional for its paradoxical combination of tolerated MDA5 hyperactivity with autoimmunity resistance.

DISCUSSION

Precise regulation of intracellular RNA duplex sensing is critical for protecting mammals against the Scylla of viral disease while also evading the Charybdis of autoimmunity. Confining sensing to only exogenous dsRNAs via PRRs such as MDA5 is therefore a bedrock feature of healthy antiviral defense. However, since MDA5 is relatively non-discriminating in its ligand preferences (mainly binding to internal segments of longer dsRNA duplexes on the order of hundreds to thousands of nucleotides in length), it has become clear in the past few years that an additional layer of control, ADAR1 p150 editing of host cellular RNAs, is at play. Nevertheless, the biological significance of the millions of A-to-I edits that modify RNA transcripts is incompletely understood.

In the same year in which he published his Nobel-awarded paper on the YF17D yellow fever vaccine that is still used today (69), virologist Max Theiler reported his isolation of Theiler’s virus and noted its similarities to poliovirus (70). Here, we identify new features of a mouse model in which we express the dsRNA-synthesizing enzyme of this picornavirus outside the viral context in the tissues of mice, providing a continuous dsRNA stimulus to the MDA5-MAVS pathway. In its strict MDA5-MAVS dependence, the model parallels the exclusively MDA5-dependent pattern for sensing of replicating picornaviruses (1, 5–8). Having previously demonstrated the lack of significant autoinflammatory “cost” to the animals, we found here that RdRp^tg mice are further able to resist an autoimmune provocation of induced SLE in the BM12 model (Fig. 1). However, the loss of a single Adar allele—which by itself is not discernibly harmful in our and others’ hands (18)—breaks the autoinflammation-protected state, resulting in a severe disease with progeric features characterized by shortened lifespan, stunted growth, premature fur graying, poorly developed teeth and skeletal abnormalities, extreme ISG elevations, and dysregulated and increased A-to-I editing. In addition to the accentuation of ISG hyper-expression compared to RdRp^tg mice, which may be causative in the pathology that emerged, the double heterozygotes hyper-expressed mRNAs for several CC and CXC cytokines (Fig. S8), a result that will be interesting to explore further in future studies of these mice. In contrast, type I IFNs and the classical proinflammatory cytokines IL6 and TNFα were not induced to higher levels by the addition of the single Adar allele knockout.

Among ADAR1 mouse models that produce autoinflammatory pathology, this is the only one with a simple haploinsufficient genotype and a wild-type remaining Adar allele. All others are either Adar^–/– (which is embryonic lethal without an Mda5 KO rescue) or utilize protein mutants such as Adar^P95A or Adar^E861A (see reference 71 for a comprehensive tabulation of existing mouse models). Relatedly, we further emphasize that the present model is nucleic acid-driven rather than MDA5 GOF mutation triggered. As such, it is a mouse model in which “endogenously” synthesized RNAs generate an autoinflammatory disease via wild-type MDA5, with the source of the dsRNA being a genomically integrated viral polymerase. Since the loss of one Adar allele is enough to tilt the immune system in RdRp^tg mice to a highly pathogenic autoinflammatory outcome, it can be speculated that, in some natural circumstances, viral infections—and the inflammatory signaling cascades they induce—have roles in triggering human autoinflammatory diseases, particularly when they also disturb the finely regulated, complex equilibrium of A-to-I editing. The RdRp^tg/– Adar^+/– model complements ADAR1 mutant mouse AGS models, which differ in utilizing mutant ADAR1 proteins (72–74) and which do not incorporate a viral polymerase. The pathological outcome in RdRp^tg/– Adar^+/– mice is variably penetrant, with about half of mice severely affected, a feature that is common in human SMS kindreds (55, 62) as well as ADAR1 mutant protein mouse models of AGS (24, 73).

In some respects, the pathologies observed in RdRp^tg/– Adar^+/– mice resemble the human interferonopathy SMS, with notable exceptions, e.g., the lack of aortic and valvular calcification typically seen in the human syndrome (54, 55). In previous studies, phenotypes resembling AGS/SMS were produced by engineering mice to express the MDA5 constitutive activation (GOF) mutants mG821S (75) or huR822Q (76). In mG821S mice, SMS-like features were observed (75); however, the model differs in that, while it recapitulates skeletal defects—also seen in RdRp^tg/– Adar^+/– mice and human SMS cases (56, 57, 59)—it did not have the dental deformities or the fur pigmentation abnormality we observed, and we did not see the deteriorating kidney function found in mG821S mice. In the picornavirus RdRp^tg model, constitutive activation of WT MDA5 expressed from the endogenous locus by upstream provision of sustained dsRNA production is insufficient, and the addition of Adar haploinsufficiency is needed. It may thus have value as a model of nucleic acid-induced autoimmunity that reflects the overwhelming majority of patients, who have normal RLR proteins. In this regard, the G821S and R822W GOF MDA5 mutant proteins do not bind dsRNA and are always “on.” This may be central to the difference. Subsequent investigation of the pathology observed in RdRtg^tg Adar haploinsufficient mice will ideally test a potential role for ZBP1, which has been identified as a main downstream effector of autoinflammation in the context of mutant ADAR1 proteins. ADAR1 acts as an upstream negative regulator of its activation, which can trigger inflammatory signaling and several varieties of regulated cell death (63, 68, 77–80)

Why ADAR1, which mediates the most abundant form of RNA editing in metazoan organisms (81), does not edit RdRp-synthesized RNAs sufficiently to prevent activation of WT MDA5 in the RdRp^tg but Adar^+/+ mouse is unclear and is an intriguing aspect to investigate in the future. The data indicate that mammals have a means to prevent viral RdRp-synthesized dsRNA from being protected (from sensing) by A-to-I editing, even when the RdRp is expressed completely outside the context of actual viral replication, i.e., without the elaborate replication factory biogenesis central to the life cycles of all positive strand RNA viruses (and also without Vpg protein capping in the case of picornaviruses). How this occurs is a key topic for investigation in the field.

In prior reports of Ifih1 gene duplication or mice carrying MDA5 GOF mutants, which also have increased expression of ISGs and relative protection against viral diseases, the mice have been shown to be more prone to either spontaneous or triggered autoimmunity (36, 42, 43, 75, 76). Thus, RdRp^tg mice may have distinctive tolerance mechanisms as evidenced by their resistance to induced SLE. A component of these tolerance mechanisms may be the increase in overall and effector regulatory T cells (Fig. 1), which are key autoimmune suppressor cells (49).

We conclude from our data that correctly regulated ADAR1 editing is a key suppressor of interferonopathic outcomes in these animals. The results also suggest that modulation of ADAR1 activity might potentially be of benefit in some human autoinflammatory diseases. Despite the loss of one allele, ADAR1 levels, specifically the levels of ADAR p150 protein, were increased equivalently in both RdRp^tg/– and RdRp^tg/– Adar^+/– animals, indicating that disease is not driven by a gene dosage effect in which low ADAR1 levels cause hypo-editing of dsRNA. Consistent with this result, a dysregulation picture emerged in which the overall number of edited sites and edited genes was increased in the double heterozygotes as compared to all other genotypes and was abnormally distributed (Fig. 7). The dysregulation could contribute to the severe phenotype observed, but again, the comparison of the degree of p150 elevation with that of RdRp^tg/– mice compels an interpretation that the level of induced ADAR1 protein is not the main driver. For example, abnormally high levels of RNA editing have been reported in SLE patients (82). When we measured the levels of proteins encoded by several uniquely edited ISG transcripts in RdRp^tg/–Adar^+/– mice, we did not identify an effect of the additional edits. We considered whether increased ISG expression might lead to increased endogenous retroelement dsRNA expression, a possibility suggested by the observed increase in editing of SINES and LINES; in this regard, Zhang et al. showed that such endogenous retroelements are, in fact, quite enriched in the 3′ UTRs of ISG mRNAs (68). Here, we found that multiple ISG mRNAs were indeed A-to-I edited, but the proportion of uniquely edited sites that were within an ISG RNA vs those of other genes was not significantly elevated.

ADAR1 activity and editing changes in autoinflammatory diseases that are not directly linked to ADAR1 mutations are still poorly understood, but our data indicate that simple expression of the enzyme at physiological or higher levels does not equate with proper function. It will be worthwhile to study changes in ADAR1 expression or dysregulation of A-to-I editing in interferonopathies and other autoimmune diseases not directly linked to ADAR1.

MATERIALS AND METHODS

Mice and nomenclature

RdRp transgenic mice on the C57/BL6J background have been previously described (37). Here, we use RdRp^tg/– to designate a mouse with one RdRp transgene allele and one normal locus lacking the chromosome 6 transgene insertion, RdRp^tg/tg for transgene homozygotes, and RdRp^tg, without italics, for the general model. Adar^+/– mice on the C57/BL6 background generated by Hartner et al. (20) were obtained from the Jackson Laboratory (MGI 3029789). There is a germline deletion of the gene region spanning exons 7–9, which results in non-functional p110 and p150 proteins (20). Mice were backcrossed for two generations onto our colony before use, and genotypes were confirmed using protocols suggested by the Jackson Laboratory. CD45.1⁺ congenic marker line mice were obtained from the Jackson Laboratory (mouse strain: 002014) and crossed onto the BM12 mouse line (Jackson Laboratory, 001162). The presence of the CD45.1 gene was confirmed via TransnetYX genotyping service (Ptprc-2 Mut probe) and confirmed with flow cytometric staining of splenocytes. BM12 genotyping was confirmed as described (45) in Ifih1 knockout mice on a B6.J background obtained from Jackson Labs (strain 015812). As much as possible, littermate controls were used for direct comparison between groups. All experiments use a mix of male and female mice for analysis. BM12 initiation of lupus-like disease was done as described (45). Briefly, 1 × 10⁸ splenocytes isolated from CD45.1⁺ BM12 in mice were injected intraperitoneally into WT (CD45.2) or RdRp^tg/tg (CD45.2) mice. Single-cell solution in PBS or PBS alone (sham) was used for injection in a 250 µL total volume. Mice were euthanized at 14 days post-injection for disease evaluation.

Animal imaging, µCT analysis, and intraocular pressure determinations

Femurs were evaluated for length, cortical bone structure, and trabecular microarchitecture using a Zeiss Xradia Versa X-ray microscope-520 (μCT, Zeiss, Dublin, CA, USA; 80 kVp, 4× objective, isotropic voxel size of 4 μm). Regions of interest were located at the femur mid-diaphysis (1 mm tall) and the distal femur metaphysis (starting at 600 μm proximal to the epiphyseal line and extending 1,000 μm proximally). Trabecular and cortical bone were segmented and analyzed using Dragonfly Pro and Bone Analysis software (Object Research Systems, Montreal, Canada) according to established guidelines (83). Trabecular bone parameters included bone volume fraction, trabecular number, trabecular thickness, and Tb.Sp. Cortical bone measures included cortical bone volume fraction (Ct.BV/TV), total bone mineral density, and cortical porosity. The bending moments of inertia (Imax and Imin), which contribute to structural stiffness under bending, were calculated. Hydrated femurs were evaluated using three-point bending to failure as described (84, 85) to determine mechanical properties (e.g., stiffness and maximum load) using a MTS Insight II benchtop tester (MTS Corp., Minneapolis, MN; 250 N load cell, 5 mm/min deflection rate, 7 mm span width) and analyzed using custom MATLAB code, Mathworks, Natick, MA, USA). Material properties (e.g., modulus and ultimate stress) were calculated using standard equations derived from engineering beam theory (84). Intraocular pressures were measured with a handheld ocular pneumotonometer (Model 30 Classic, Medtronic) as described (86).

ETHICS APPROVAL

All mouse experiments were performed in accordance with the Institutional Animal Care and Use Committee (IACUC) guidelines and approved procedures for the University of Colorado, Anschutz Medical Center (IACUC protocol no. 00116).

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/mbio.01492-25.

Supplemental Material. mbio.01492-25-s0001.pdf.

Fig. S1–S10; supplemental methods.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.