Ligand-dependent hedgehog signaling maintains an undifferentiated, malignant osteosarcoma phenotype.

TP53 and RB1 loss-of-function mutations are common in osteosarcoma. During development, combined loss of TP53 and RB1 function leads to downregulation of autophagy and the aberrant formation of primary cilia, cellular organelles essential for the transmission of canonical Hedgehog (Hh) signaling. Excess cilia formation then leads to hypersensitivity to Hedgehog (Hh) ligand signaling. In mouse and human models, we now show that osteosarcomas with mutations in TP53 and RB1 exhibit enhanced ligand-dependent Hh pathway activation through Smoothened (SMO), a transmembrane signaling molecule required for activation of the canonical Hh pathway. This dependence is mediated by hypersensitivity to Hh ligand and is accompanied by impaired autophagy and increased primary cilia formation and expression of Hh ligand in vivo. Using a conditional genetic mouse model of Trp53 and Rb1 inactivation in osteoblast progenitors, we further show that deletion of Smo converts the highly malignant osteosarcoma phenotype to benign, well differentiated bone tumors. Conversely, conditional overexpression of SHH ligand, or a gain-of-function SMO mutant in committed osteoblast progenitors during development blocks terminal bone differentiation. Finally, we demonstrate that the SMO antagonist sonidegib (LDE225) induces growth arrest and terminal differentiation in vivo in osteosarcomas that express primary cilia and Hh ligand combined with mutations in TP53. These results provide a mechanistic framework for aberrant Hh signaling in osteosarcoma based on defining mutations in the tumor suppressor, TP53.

ADAR1p150 prevents MDA5 and PKR activation via distinct mechanisms to avert fatal autoinflammation.

Effective immunity requires the innate immune system to distinguish foreign nucleic acids from cellular ones. Cellular double-stranded RNAs (dsRNAs) are edited by the RNA-editing enzyme ADAR1 to evade being recognized as viral dsRNA by cytoplasmic dsRNA sensors, including MDA5 and PKR. The loss of ADAR1-mediated RNA editing of cellular dsRNA activates MDA5. Additional RNA-editing-independent functions of ADAR1 have been proposed, but a specific mechanism has not been delineated. We now demonstrate that the loss of ADAR1-mediated RNA editing specifically activates MDA5, whereas loss of the cytoplasmic ADAR1p150 isoform or its dsRNA-binding activity enabled PKR activation. Deleting both MDA5 and PKR resulted in complete rescue of the embryonic lethality of Adar1p150-/- mice to adulthood, contrasting with the limited or no rescue by removing MDA5 or PKR alone. Our findings demonstrate that MDA5 and PKR are the primary in vivo effectors of fatal autoinflammation following the loss of ADAR1p150.

Generation of a new Adar1p150 -/- mouse demonstrates isoform-specific roles in embryonic development and adult homeostasis.

The RNA editing enzyme adenosine deaminase acting on RNA 1 (ADAR1) is an essential regulator of the innate immune response to both cellular and viral double-stranded RNA (dsRNA). Adenosine-to-inosine (A-to-I) editing by ADAR1 modifies the sequence and structure of endogenous dsRNA and masks it from the cytoplasmic dsRNA sensor melanoma differentiation-associated protein 5 (MDA5), preventing innate immune activation. Loss-of-function mutations in ADAR are associated with rare autoinflammatory disorders including Aicardi-Goutières syndrome (AGS), defined by a constitutive systemic up-regulation of type I interferon (IFN). The murine Adar gene encodes two protein isoforms with distinct functions: ADAR1p110 is constitutively expressed and localizes to the nucleus, whereas ADAR1p150 is primarily cytoplasmic and is inducible by IFN. Recent studies have demonstrated the critical requirement for ADAR1p150 to suppress innate immune activation by self dsRNAs. However, detailed in vivo characterization of the role of ADAR1p150 during development and in adult mice is lacking. We identified a new ADAR1p150-specific knockout mouse mutant based on a single nucleotide deletion that resulted in the loss of the ADAR1p150 protein without affecting ADAR1p110 expression. The Adar1p150 -/- died embryonically at E11.5-E12.5 accompanied by cell death in the fetal liver and an activated IFN response. Somatic loss of ADAR1p150 in adults was lethal and caused rapid hematopoietic failure, demonstrating an ongoing requirement for ADAR1p150 in vivo. The generation and characterization of this mouse model demonstrates the essential role of ADAR1p150 in vivo and provides a new tool for dissecting the functional differences between ADAR1 isoforms and their physiological contributions.

Over-expression of ADAR1 in mice does not initiate or accelerate cancer formation in vivo.

Adenosine to inosine editing (A-to-I) in regions of double stranded RNA (dsRNA) is mediated by adenosine deaminase acting on RNA 1 (ADAR1) or ADAR2. ADAR1 and A-to-I editing levels are increased in many human cancers. Inhibition of ADAR1 has emerged as a high priority oncology target, however, whether ADAR1 overexpression enables cancer initiation or progression has not been directly tested. We established a series of in vivo models to allow overexpression of full-length ADAR1, or its individual isoforms, to test if increased ADAR1 expression was oncogenic. Widespread over-expression of ADAR1 or the p110 or p150 isoforms individually as sole lesions was well tolerated and did not result in cancer initiation. Therefore, ADAR1 overexpression alone is not sufficient to initiate cancer. We demonstrate that endogenous ADAR1 and A-to-I editing increased upon immortalization in murine cells, consistent with the observations from human cancers. We tested if ADAR1 over-expression could co-operate with cancer initiated by loss of tumour suppressors using a model of osteosarcoma. We did not see a disease potentiating or modifying effect of overexpressing ADAR1 or its isoforms in the models assessed. We conclude that increased ADAR1 expression and A-to-I editing in cancers is most likely a consequence of tumor formation.

The phenotype of the most common human ADAR1p150 Zα mutation P193A in mice is partially penetrant.

ADAR1 -mediated A-to-I RNA editing is a self-/non-self-discrimination mechanism for cellular double-stranded RNAs. ADAR mutations are one cause of Aicardi-Goutières Syndrome, an inherited paediatric encephalopathy, classed as a "Type I interferonopathy." The most common ADAR1 mutation is a proline 193 alanine (p.P193A) mutation, mapping to the ADAR1p150 isoform-specific Zα domain. Here, we report the development of an independent murine P195A knock-in mouse, homologous to human P193A. The Adar1P195A/P195A mice are largely normal and the mutation is well tolerated. When the P195A mutation is compounded with an Adar1 null allele (Adar1P195A/- ), approximately half the animals are runted with a shortened lifespan while the remaining Adar1P195A/- animals are normal, contrasting with previous reports. The phenotype of the Adar1P195A/- animals is both associated with the parental genotype and partly non-genetic/environmental. Complementation with an editing-deficient ADAR1 (Adar1P195A/E861A ), or the loss of MDA5, rescues phenotypes in the Adar1P195A/- mice.

ADAR1p150 Prevents MDA5 and PKR Activation via Distinct Mechanisms to Avert Fatal Autoinflammation.

Effective immunity requires the innate immune system to distinguish foreign (non-self) nucleic acids from cellular (self) nucleic acids. Cellular double-stranded RNAs (dsRNAs) are edited by the RNA editing enzyme ADAR1 to prevent their dsRNA structure pattern being recognized as viral dsRNA by cytoplasmic dsRNA sensors including MDA5, PKR and ZBP1. A loss of ADAR1-mediated RNA editing of cellular dsRNA activates MDA5. However, additional RNA editing-independent functions of ADAR1 have been proposed, but a specific mechanism has not been delineated. We now demonstrate that the loss of ADAR1-mediated RNA editing specifically activates MDA5, while loss of the cytoplasmic ADAR1p150 isoform or its dsRNA binding activity enabled PKR activation. Deleting both MDA5 and PKR resulted in complete rescue of the embryonic lethality of Adar1p150 -/- mice to adulthood, contrasting with the limited or no rescue by removing MDA5, PKR or ZBP1 alone, demonstrating that this is a species conserved function of ADAR1p150. Our findings demonstrate that MDA5 and PKR are the primary in vivo effectors of fatal autoinflammation following the loss of ADAR1p150.

Ablation of Adar1 in myeloid cells imprints a global antiviral state in the lung and heightens early immunity against SARS-CoV-2.

Under normal homeostatic conditions, self-double-stranded RNA (self-dsRNA) is modified by adenosine deaminase acting on RNA 1 (ADAR1) to prevent the induction of a type I interferon-mediated inflammatory cascade. Antigen-presenting cells (APCs) sense pathogen-associated molecular patterns, such as dsRNA, to activate the immune response. The impact of ADAR1 on the function of APCs and the consequences to immunity are poorly understood. Here, we show that ADAR1 deletion in CD11c+ APCs leads to (1) a skewed myeloid cell compartment enriched in inflammatory cDC2-like cells, (2) enhanced numbers of activated tissue resident memory T cells in the lung, and (3) the imprinting of a broad antiviral transcriptional signature across both immune and non-immune cells. The resulting changes can be partially reversed by blocking IFNAR1 signaling and promote early resistance against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. Our study provides insight into the consequences of self-dsRNA sensing in APCs on the immune system.

Srsf2P95H/+ co-operates with loss of TET2 to promote myeloid bias and initiate a chronic myelomonocytic leukemia-like disease in mice.

Recurrent mutations in RNA splicing proteins and epigenetic regulators contribute to the development of myelodysplastic syndrome (MDS) and related myeloid neoplasms. In chronic myelomonocytic leukemia (CMML), SRSF2 mutations occur in ~50% of patients and TET2 mutations in ~60%. Clonal analysis indicates that either mutation can arise as the founder lesion. Based on human cancer genetics we crossed an inducible Srsf2P95H/+ mutant model with Tet2fl/fl mice to mutate both concomitantly in hematopoietic stem cells. At 20-24 weeks post mutation induction, we observed subtle differences in the Srsf2/Tet2 mutants compared to either single mutant. Under conditions of native hematopoiesis with aging, we see a distinct myeloid bias and monocytosis in the Srsf2/Tet2 mutants. A subset of the compound Srsf2/Tet2 mutants display an increased granulocytic and distinctive monocytic proliferation (myelomonocytic hyperplasia), with increased immature promonocytes and monoblasts and binucleate promonocytes. Exome analysis of progressed disease demonstrated mutations in genes and pathways similar to those reported in human CMML. Upon transplantation, recipients developed leukocytosis, monocytosis, and splenomegaly. We reproduce Srsf2/Tet2 co-operativity in vivo, yielding a disease with core characteristics of CMML, unlike single Srsf2 or Tet2 mutation. This model represents a significant step toward building high fidelity and genetically tractable models of CMML.

Direct identification of A-to-I editing sites with nanopore native RNA sequencing.

Inosine is a prevalent RNA modification in animals and is formed when an adenosine is deaminated by the ADAR family of enzymes. Traditionally, inosines are identified indirectly as variants from Illumina RNA-sequencing data because they are interpreted as guanosines by cellular machineries. However, this indirect method performs poorly in protein-coding regions where exons are typically short, in non-model organisms with sparsely annotated single-nucleotide polymorphisms, or in disease contexts where unknown DNA mutations are pervasive. Here, we show that Oxford Nanopore direct RNA sequencing can be used to identify inosine-containing sites in native transcriptomes with high accuracy. We trained convolutional neural network models to distinguish inosine from adenosine and guanosine, and to estimate the modification rate at each editing site. Furthermore, we demonstrated their utility on the transcriptomes of human, mouse and Xenopus. Our approach expands the toolkit for studying adenosine-to-inosine editing and can be further extended to investigate other RNA modifications.

ADAR1 masks the cancer immunotherapeutic promise of ZBP1-driven necroptosis.

Only a small proportion of patients with cancer show lasting responses to immune checkpoint blockade (ICB)-based monotherapies. The RNA-editing enzyme ADAR1 is an emerging determinant of resistance to ICB therapy and prevents ICB responsiveness by repressing immunogenic double-stranded RNAs (dsRNAs), such as those arising from the dysregulated expression of endogenous retroviral elements (EREs)1-4. These dsRNAs trigger an interferon-dependent antitumour response by activating A-form dsRNA (A-RNA)-sensing proteins such as MDA-5 and PKR5. Here we show that ADAR1 also prevents the accrual of endogenous Z-form dsRNA elements (Z-RNAs), which were enriched in the 3' untranslated regions of interferon-stimulated mRNAs. Depletion or mutation of ADAR1 resulted in Z-RNA accumulation and activation of the Z-RNA sensor ZBP1, which culminated in RIPK3-mediated necroptosis. As no clinically viable ADAR1 inhibitors currently exist, we searched for a compound that can override the requirement for ADAR1 inhibition and directly activate ZBP1. We identified a small molecule, the curaxin CBL0137, which potently activates ZBP1 by triggering Z-DNA formation in cells. CBL0137 induced ZBP1-dependent necroptosis in cancer-associated fibroblasts and reversed ICB unresponsiveness in mouse models of melanoma. Collectively, these results demonstrate that ADAR1 represses endogenous Z-RNAs and identifies ZBP1-mediated necroptosis as a new determinant of tumour immunogenicity masked by ADAR1. Therapeutic activation of ZBP1-induced necroptosis provides a readily translatable avenue for rekindling the immune responsiveness of ICB-resistant human cancers.

Genome-wide screening identifies cell-cycle control as a synthetic lethal pathway with SRSF2P95H mutation.

Current strategies to target RNA splicing mutant myeloid cancers proposes targeting the remaining splicing apparatus. This approach has only been modestly sensitizing and is also toxic to non-mutant-bearing wild-type cells. To explore potentially exploitable genetic interactions with spliceosome mutations, we combined data mining and functional screening for synthetic lethal interactions with an Srsf2P95H/+ mutation. Analysis of missplicing events in a series of both human and murine SRSF2P95H mutant samples across multiple myeloid diseases (acute myeloid leukemia, myelodysplastic syndromes, chronic myelomonocytic leukemia) was performed to identify conserved missplicing events. From this analysis, we identified that the cell-cycle and DNA repair pathways were overrepresented within the conserved misspliced transcript sets. In parallel, to functionally define pathways essential for survival and proliferation of Srsf2P95H/+ cells, we performed a genome-wide Clustered regularly interspaced short palindromic repeat loss-of-function screen using Hoxb8 immortalized R26-CreERki/+Srsf2P95H/+ and R26-CreERki/+Srsf2+/+ cell lines. We assessed loss of single guide RNA representation at 3 timepoints: immediately after Srsf2P95H/+ activation, and at 1 week and 2 weeks after Srsf2P95H/+ mutation. Pathway analysis demonstrated that the cell-cycle and DNA damage response pathways were among the top synthetic lethal pathways with Srsf2P95H/+ mutation. Based on the loss of guide RNAs targeting Cdk6, we identified that palbociclib, a CDK6 inhibitor, showed preferential sensitivity in Srsf2P95H/+ cell lines and in primary nonimmortalized lin-cKIT+Sca-1+ cells compared with wild-type controls. Our data strongly suggest that the cell-cycle and DNA damage response pathways are required for Srsf2P95H/+ cell survival, and that palbociclib could be an alternative therapeutic option for targeting SRSF2 mutant cancers.

Rothmund-Thomson Syndrome-Like RECQL4 Truncating Mutations Cause a Haploinsufficient Low-Bone-Mass Phenotype in Mice.

Rothmund-Thomson syndrome (RTS) is an autosomal recessive disorder characterized by defects in the skeletal system, such as bone hypoplasia, short stature, low bone mass, and an increased incidence of osteosarcoma. RTS type 2 patients have germ line compound biallelic protein-truncating mutations of RECQL4. As existing murine models employ Recql4 null alleles, we have attempted to more accurately model RTS by generating mice with patient-mimicking truncating Recql4 mutations. Truncating mutations impaired the stability and subcellular localization of RECQL4 and resulted in homozygous embryonic lethality and a haploinsufficient low-bone mass phenotype. Combination of a truncating mutation with a conditional Recql4 null allele demonstrated that the skeletal defects were intrinsic to the osteoblast lineage. However, the truncating mutations did not promote tumorigenesis. We utilized murine Recql4 null cells to assess the impact of human RECQL4 mutations using an in vitro complementation assay. While some mutations created unstable protein products, others altered subcellular localization of the protein. Interestingly, the severity of the phenotypes correlated with the extent of protein truncation. Collectively, our results reveal that truncating RECQL4 mutations in mice lead to an osteoporosis-like phenotype through defects in early osteoblast progenitors and identify RECQL4 gene dosage as a novel regulator of bone mass.

What do editors do? Understanding the physiological functions of A-to-I RNA editing by adenosine deaminase acting on RNAs.

Adenosine-to-inosine (A-to-I) editing is a post-transcriptional modification of RNA which changes its sequence, coding potential and secondary structure. Catalysed by the adenosine deaminase acting on RNA (ADAR) proteins, ADAR1 and ADAR2, A-to-I editing occurs at approximately 50 000-150 000 sites in mice and into the millions of sites in humans. The vast majority of A-to-I editing occurs in repetitive elements, accounting for the discrepancy in total numbers of sites between species. The species-conserved primary role of editing by ADAR1 in mammals is to suppress innate immune activation by unedited cell-derived endogenous RNA. In the absence of editing, inverted paired sequences, such as Alu elements, are thought to form stable double-stranded RNA (dsRNA) structures which trigger activation of dsRNA sensors, such as MDA5. A small subset of editing sites are within coding sequences and are evolutionarily conserved across metazoans. Editing by ADAR2 has been demonstrated to be physiologically important for recoding of neurotransmitter receptors in the brain. Furthermore, changes in RNA editing are associated with various pathological states, from the severe autoimmune disease Aicardi-Goutières syndrome, to various neurodevelopmental and psychiatric conditions and cancer. However, does detection of an editing site imply functional importance? Genetic studies in humans and genetically modified mouse models together with evolutionary genomics have begun to clarify the roles of A-to-I editing in vivo. Furthermore, recent developments suggest there may be the potential for distinct functions of editing during pathological conditions such as cancer.

Enhancing mitochondrial function in vivo rescues MDS-like anemia induced by pRb deficiency.

Erythropoiesis is intimately coupled to cell division, and deletion of the cell cycle regulator retinoblastoma protein (pRb) causes anemia in mice. Erythroid-specific deletion of pRb has been found to result in inefficient erythropoiesis because of deregulated coordination of cell cycle exit and mitochondrial biogenesis. However, the pathophysiology remains to be fully described, and further characterization of the link between cell cycle regulation and mitochondrial function is needed. To this end we further assessed conditional erythroid-specific deletion of pRb. This resulted in macrocytic anemia, despite elevated levels of erythropoietin (Epo), and an accumulation of erythroid progenitors in the bone marrow, a phenotype strongly resembling refractory anemia associated with myelodysplastic syndromes (MDS). Using high-fractionation fluorescence-activated cell sorting analysis for improved phenotypic characterization, we illustrate that erythroid differentiation was disrupted at the orthochromatic stage. Transcriptional profiling of sequential purified populations revealed failure to upregulate genes critical for mitochondrial function such as Pgc1ß, Alas2, and Abcb7 specifically at the block, together with disturbed heme production and iron transport. Notably, deregulated ABCB7 causes ring sideroblastic anemia in MDS patients, and the mitochondrial co-activator PGC1ß is heterozygously lost in del5q MDS. Importantly, the anemia could be rescued through enhanced PPAR signaling in vivo via either overexpression of Pgc1ß or bezafibrate administration. In conclusion, lack of pRb results in MDS-like anemia with disrupted differentiation and impaired mitochondrial function at the orthochromatic erythroblast stage. Our findings reveal for the first time a role for pRb in heme and iron regulation, and indicate that pRb-induced anemia can be rescued in vivo through therapeutic enhancement of PPAR signaling.

Hematopoietic stem and progenitor cell-restricted Cdx2 expression induces transformation to myelodysplasia and acute leukemia.

The caudal-related homeobox transcription factor CDX2 is expressed in leukemic cells but not during normal blood formation. Retroviral overexpression of Cdx2 induces AML in mice, however the developmental stage at which CDX2 exerts its effect is unknown. We developed a conditionally inducible Cdx2 mouse model to determine the effects of in vivo, inducible Cdx2 expression in hematopoietic stem and progenitor cells (HSPCs). Cdx2-transgenic mice develop myelodysplastic syndrome with progression to acute leukemia associated with acquisition of additional driver mutations. Cdx2-expressing HSPCs demonstrate enrichment of hematopoietic-specific enhancers associated with pro-differentiation transcription factors. Furthermore, treatment of Cdx2 AML with azacitidine decreases leukemic burden. Extended scheduling of low-dose azacitidine shows greater efficacy in comparison to intermittent higher-dose azacitidine, linked to more specific epigenetic modulation. Conditional Cdx2 expression in HSPCs is an inducible model of de novo leukemic transformation and can be used to optimize treatment in high-risk AML.

Publisher Correction: Dynamic regulation of Z-DNA in the mouse prefrontal cortex by the RNA-editing enzyme Adar1 is required for fear extinction.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

ADAR1-Dependent RNA Editing Promotes MET and iPSC Reprogramming by Alleviating ER Stress.

RNA editing of adenosine to inosine (A to I) is catalyzed by ADAR1 and dramatically alters the cellular transcriptome, although its functional roles in somatic cell reprogramming are largely unexplored. Here, we show that loss of ADAR1-mediated A-to-I editing disrupts mesenchymal-to-epithelial transition (MET) during induced pluripotent stem cell (iPSC) reprogramming and impedes acquisition of induced pluripotency. Using chemical and genetic approaches, we show that absence of ADAR1-dependent RNA editing induces aberrant innate immune responses through the double-stranded RNA (dsRNA) sensor MDA5, unleashing endoplasmic reticulum (ER) stress and hindering epithelial fate acquisition. We found that A-to-I editing impedes MDA5 sensing and sequestration of dsRNAs encoding membrane proteins, which promote ER homeostasis by activating the PERK-dependent unfolded protein response pathway to consequently facilitate MET. This study therefore establishes a critical role for ADAR1 and its A-to-I editing activity during cell fate transitions and delineates a key regulatory layer underlying MET to control efficient reprogramming.

Dynamic regulation of Z-DNA in the mouse prefrontal cortex by the RNA-editing enzyme Adar1 is required for fear extinction.

DNA forms conformational states beyond the right-handed double helix; however, the functional relevance of these noncanonical structures in the brain remains unknown. Here we show that, in the prefrontal cortex of mice, the formation of one such structure, Z-DNA, is involved in the regulation of extinction memory. Z-DNA is formed during fear learning and reduced during extinction learning, which is mediated, in part, by a direct interaction between Z-DNA and the RNA-editing enzyme Adar1. Adar1 binds to Z-DNA during fear extinction learning, which leads to a reduction in Z-DNA at sites where Adar1 is recruited. Knockdown of Adar1 leads to an inability to modify a previously acquired fear memory and blocks activity-dependent changes in DNA structure and RNA state-effects that are fully rescued by the introduction of full-length Adar1. These findings suggest a new mechanism of learning-induced gene regulation that is dependent on proteins that recognize alternate DNA structure states, which are required for memory flexibility.

The majority of A-to-I RNA editing is not required for mammalian homeostasis.

BACKGROUND: Adenosine-to-inosine (A-to-I) RNA editing, mediated by ADAR1 and ADAR2, occurs at tens of thousands to millions of sites across mammalian transcriptomes. A-to-I editing can change the protein coding potential of a transcript and alter RNA splicing, miRNA biology, RNA secondary structure and formation of other RNA species. In vivo, the editing-dependent protein recoding of GRIA2 is the essential function of ADAR2, while ADAR1 editing prevents innate immune sensing of endogenous RNAs by MDA5 in both human and mouse. However, a significant proportion of A-to-I editing sites can be edited by both ADAR1 and ADAR2, particularly within the brain where both are highly expressed. The physiological function(s) of these shared sites, including those evolutionarily conserved, is largely unknown. RESULTS: To generate completely A-to-I editing-deficient mammals, we crossed the viable rescued ADAR1-editing-deficient animals (Adar1E861A/E861AIfih1-/-) with rescued ADAR2-deficient (Adarb1-/-Gria2R/R) animals. Unexpectedly, the global absence of editing was well tolerated. Adar1E861A/E861AIfih1-/-Adarb1-/-Gria2R/R were recovered at Mendelian ratios and age normally. Detailed transcriptome analysis demonstrated that editing was absent in the brains of the compound mutants and that ADAR1 and ADAR2 have similar editing site preferences and patterns. CONCLUSIONS: We conclude that ADAR1 and ADAR2 are non-redundant and do not compensate for each other's essential functions in vivo. Physiologically essential A-to-I editing comprises a small subset of the editome, and the majority of editing is dispensable for mammalian homeostasis. Moreover, in vivo biologically essential protein recoding mediated by A-to-I editing is an exception in mammals.