Abstract
Adenosine deaminase acting on RNA (ADAR)-mediated RNA editing has emerged as a powerful and precise technology for modifying RNA transcripts, enabling correction of disease-causing mutations without permanent changes to the genome. Recent advances in ADAR protein engineering, guide RNA design, and delivery methods have significantly improved editing efficiency and specificity, overcoming many initial limitations. These developments have expanded the therapeutic potential of ADAR-based editing across a range of conditions, including genetic disorders, cancer, metabolic diseases, and neurodegenerative disorders. Notably, several ADAR-based therapeutics have now entered early clinical trials, marking a critical milestone in translating this technology from bench to bedside. Moreover, its inherent programmability, reversibility, and transient nature make ADAR-mediated RNA editing a highly attractive platform for personalized medicine, enabling tailored interventions based on individual genetic profiles and disease contexts. This review provides a comprehensive comparison of recent innovative advancements in ADAR-based RNA editing technologies, their use in diverse contexts pertinent to human diseases, the key challenges that remain, and future directions for their therapeutic implementation.
Keywords
Introduction
Over the past decade, there has been a tremendous expansion of technologies for genomic and transcriptomic engineering. While the field of genome editing has been revolutionized by the development of innovative tools like clustered regularly interspaced short palindromic repeats (CRISPR)–Cas9, which allows precise and efficient modifications to DNA, there has been growing momentum for a field that is similarly transformative, RNA editing, which involves post-transcriptional alterations of RNA nucleotide sequence. RNA editing holds great therapeutic potential due to its ability to modify specific target sites in a reversible and tunable manner, distinguishing it from DNA editing, which causes nontemporary changes to the genome and can introduce unintended errors with potentially significant and permanent consequences.1,2 The transient and controllable nature of RNA editing makes it particularly suitable for treating genetic diseases that do not necessitate stable correction or that are particularly vulnerable to the consequences of inadvertent errors. 3 Given the limitations presented by traditional gene therapy, including immune responses, delivery challenges, and low efficacy, RNA editing may represent a promising alternative, offering a safer and more flexible way to address diseases with genetic alterations and/or dysregulated protein expression or function. 4
Another notable advantage of RNA editing is its potential applicability to the treatment of polygenic and complex diseases, which are not readily amenable to DNA-based gene therapy. 5 Recent endeavors have led to the development of platforms that enable simultaneous editing at multiple sites and different bases,6–8 demonstrating the feasibility of multiplexed RNA editing and unlocking its potential for intervention of diseases where multiple genes and pathways are involved. Considering the higher prevalence of polygenic and complex diseases compared to monogenic disorders, extending the application of RNA editing to such challenging conditions offers potential to benefit a broader range of patient populations. 9
The majority of human genetic diseases, only 10% of which can be addressed by conventional treatments, are attributable to single-nucleotide variants (SNVs).1,10 This most common type of variation in the human genome accounts for approximately 90% of all genetic alterations in ClinVar, 11 a public archive of human genetic variations with information on their phenotypic and clinical relevance, 23% of which are pathogenic or likely pathogenic. 12 With RNA editing primarily targeting single nucleotides and capable of correcting many known pathogenic SNVs, 10 it is poised to become a crucial part of the next generation of RNA therapeutics.
In recent years, there has been a surge in the development of tools allowing site-directed RNA editing (SDRE), which involves targeted modification of specific nucleotides, such as adenosine-to-inosine (A-to-I) and cytosine-to-uridine (C-to-U), at defined sites. 13 These programmable editing systems harness RNA editing enzymes, including adenosine deaminase acting on RNA (ADAR) and apolipoprotein B messenger RNA (mRNA) editing catalytic polypeptide-like (APOBEC) enzyme families, which mediate A-to-I and C-to-U conversions, respectively. 13 Notably, the ADAR-based platforms have drawn considerable attention as a therapeutic modality owing to their ability to mediate precise edits with low genotoxicity and immunogenicity risks. 5 In these systems, endogenous or exogenously delivered ADARs are directed to specific RNA transcripts by a guide RNA (gRNA) that is complementary to the target sequence with a C mismatch opposite to the target A (Fig. 1). 14 The double-stranded RNA (dsRNA) formed by the gRNA and the target strand serves as a substrate for hydrolytic deamination of adenosine to inosine by the ADAR enzymes. 14 The resulting inosine is recognized as guanosine (G) by the translational and splicing machineries, potentially leading to recoding of transcripts and changes in splice site recognition, RNA stability, and protein interactions 4 —effects that can have functional consequences at the protein level and be leveraged for therapeutic purposes.
Overview of ADAR-based RNA editing systems and their disease applications. The top section lists endogenous (left) and exogenous (right) ADAR-based editing systems. The bottom section highlights diseases in which these systems have been experimentally applied. ADAR, adenosine deaminase acting on RNA; AIMers, A-to-I RNA base editing oligonucleotides; A-to-I, adenosine-to-inosine; cadRNAs, circular ADAR-recruiting guide RNAs; CIRTS, CRISPR–Cas-inspired RNA targeting system; CLUSTER, clustered ADAR-recruiting guide RNAs for efficient RNA editing; CRISPR, clustered regularly interspaced short palindromic repeats; gRNA, guide ribonucleic acid; LEAPER, leveraging endogenous ADAR for programmable editing of RNA; MIRROR, mimicking inverted repeats to recruit ADARs using engineered oligoribonucleotides; mRNA, messenger ribonucleic acid; REPAIR, RNA editing for programmable A to I replacement; RESCUE, RNA editing for specific C-to-U exchange; RESTORE, recruiting endogenous ADAR to specific transcripts for oligonucleotide-mediated RNA editing; RNA, ribonucleic acid.
This review aims to provide an overview of recent developments in ADAR-based RNA editing approaches and their therapeutic applications, and discuss the challenges and future prospects for their integration into next-generation RNA therapeutics.
ADAR Enzymes: Isoforms, Structure, and Mechanisms of Editing
ADAR family proteins and their structure
The mammalian genome encodes three members of the ADAR family: ADAR1, ADAR2, and ADAR3, also known as ADAR, ADARB1, and ADARB2, respectively. 14 For ADAR1, there are two isoforms named based on their molecular mass, including ADAR1p150 (150 kDa) and ADAR1p110 (110 kDa). 15 ADAR2 exists in two major isoforms: the shorter ADAR2a and the longer ADAR2b, referred to as ADAR2S and ADAR2L, respectively. 16 While all isoforms of ADAR1 and ADAR2 are catalytically active, ADAR3 lacks deaminase activity, although it plays a role in the regulation of RNA editing through competitive inhibition of ADAR1 and ADAR2.3,17–19
All ADARs share a modular structure comprising a carboxy-terminal deaminase domain (DD) and a variable number of double-stranded RNA-binding domains (dsRBDs), with ADAR1 containing three dsRBDs, and ADAR2 and ADAR3 each harboring two dsRBDs (Fig. 2).14,20 The dsRBD, which is approximately 65–70 amino acids in length, recognizes and binds dsRNA, whether the substrate is perfect duplex RNA or an imperfect structure containing bulges, hairpins, and mismatches.14,21–24 The DD contains an active site for the deamination of A to I, enabling catalytic activity in ADAR1 and ADAR2, but not in ADAR3, which has an amino acid substitution at a critical residue in its DD.25,26 In ADAR2b, an Alu insertion is uniquely present within the DD, which may explain its reduced activity compared to ADAR2a.3,27,28 In addition to the DD and dsRBDs, ADAR1 contains Z-DNA/RNA-binding domains (ZBDs) at its amino-terminus with ADAR1p150 and ADAR1p110 comprising two ZBDs (Zα and Zβ) and one ZBD (Zβ), respectively (Fig. 2).20,29 The Zα and Zβ domains differ in their nucleic acid-binding properties; Zα binds both Z-DNA and Z-RNA, whereas Zβ lacks this capability. 30 It has been shown that the catalytic domain of ADAR1 alone exhibits editing activity, indicating that the two ZBDs are not strictly required for RNA editing. 31 However, the Zα domain is necessary for editing a subset of RNA substrates and plays a role in ADAR1 localization and editing efficiency.32–34 Within the Zα domain of ADAR1p150, there is also a unique nuclear export signal (NES), which, together with a nuclear localization signal (NLS) in the third dsRBD, enables ADAR1p150 to shuttle between the nucleus and cytoplasm. 14 It is indeed present in both compartments, but its expression, which is driven by an interferon (IFN)-inducible promoter, is detected predominantly in the cytoplasm, whereas ADAR1p110 and ADAR2 are primarily localized to the nucleus.14,35,36 The structural properties of ADARs are closely related to their substrate specificity and binding affinity,3,37–39 which must be taken into consideration for the design of effective and precise gRNAs tailored to specific isoforms.
Domain structure of human ADAR enzymes. All proteins are oriented from the amino (N) terminus (left) to the carboxy (C) terminus (right). Structural domains are indicated by different colors, as defined in the key at the bottom of the figure. Z-DNA/RNA-binding domains are present only in ADAR1. The two ADAR1 isoforms (ADAR1p110 and ADAR1p150) are distinguished by the presence or absence of the Zα domain and the nuclear export signal (NES). The two ADAR2 isoforms (ADAR2a and ADAR2b) are differentiated by the presence or absence of an Alu insertion within the deaminase domain. ADAR3, which is catalytically inactive, contains an arginine-rich (R) motif at its N-terminus. All isoforms include a nuclear localization signal (NLS). ADAR, adenosine deaminase acting on RNA; dsRNA, double-stranded ribonucleic acid.
ADAR expression
Successful therapeutic applications of the ADAR-based editing systems, particularly those leveraging endogenous enzymes, depend on the expression of ADARs in target tissues. Among the three ADAR family members, ADAR1 is overall the most highly and broadly expressed across tissues with its two isoforms exhibiting distinct expression patterns.40–42 ADAR1p110 is known to be expressed in a constitutive and ubiquitous manner and to be responsible for the majority of editing activity, whereas ADAR1p150 expression is induced by IFN or in response to viral infection.29,36,43,44 Previous studies took advantage of this inducible effect of IFN on ADAR1p150 to enhance RNA editing efficiency and successfully achieved higher editing yield in vitro.45,46 However, in vivo, despite the ubiquitous expression of ADAR1p150, its editing activity has been documented as minimal. For example, Hood et al. have shown that the virus-mediated induction of ADAR1p150 in brain regions does not lead to significant changes in site-specific A-to-I conversion, 44 and Kim et al. have reported that fewer than 2% of RNA editing sites are preserved in the brain of Adar1 p110/Adar2 double knockout mice, 47 suggesting a limited role of ADAR1p150 in RNA editing in vivo. However, certain sites are uniquely targeted by ADAR1p150, and editing those ADAR1p150-specific sites is required for the prevention of melanoma differentiation-associated protein 5 (MDA5) sensing of endogenous dsRNA as nonself, thereby suppressing innate immune activation, 47 which highlights the indispensable function of ADAR1p150 in the context of RNA editing and immune response.
ADAR2, like ADAR1p110, is constitutively and ubiquitously expressed, but its expression is particularly abundant in the brain, lung, bladder, and arteries.16,48,49 In ADAR2 knockout mice, which experience seizures and postnatal death, the lethal phenotype was rescued through exonic substitution of an underedited transcript of the glutamate ionotropic receptor AMPA type subunit 2 (GRIA2) gene with an edited version at the so-called Q/R site—where a glutamine (Q) codon is changed to an arginine (R) codon through A-to-I editing—underscoring the role of ADAR2 for site-specific editing.3,50
The expression of ADAR3 is restricted to the brain, distinguishing it from the ubiquitous expression of ADAR1 and ADAR2, and negatively correlates with editing.3,17 ADAR3, upon its overexpression in vitro, results in inhibition of the GRIA2 Q/R site editing via direct binding to GRIA2 in competition with ADAR2, 18 supporting an inverse correlation between ADAR3 expression and editing levels at specific sites. In line with this finding, the study by Raghava Kurup et al. reported a correlation between ADAR3 binding and reduced editing at over 400 sites in the glioblastoma transcriptome, 19 which further indicates a negative regulatory role for ADAR3 in A-to-I editing. It is important to note, however, that the deficiency of Adar3 in vivo does not substantially alter overall RNA editing, 51 suggesting that ADAR3 function may be substrate-specific or dependent on specific conditions or contexts.
The ubiquitous expression of ADARs, especially ADAR1 and ADAR2, creates opportunities for the development of many therapeutic applications. However, given the variability of their expression levels across cell types and tissues, the feasibility of RNA editing via endogenous ADAR recruitment needs to be evaluated carefully and thoroughly in each relevant cellular or tissue context. In the study by Schaffer et al., where global A-to-I RNA editing profiles were examined across over 1,000 human cell lines, it has also been shown that tissue-dependent variability in editing profiles is lost in cell lines and the editing levels in cell lines are generally lower compared to those in the corresponding tissues, 52 indicating that the editing environment in vitro differs from that in vivo, and it is critical to select physiologically relevant models for assessing in vivo activity of therapeutic ADAR-based platforms. In addition, according to the A-to-I editing profiling study of 8,551 human samples representing over 50 body sites from the Genotype-Tissue Expression project, there is only a moderate correlation between the mRNA expression of ADAR1 and ADAR2 and A-to-I editing, suggesting that editing is regulated by additional factors. 53 Future studies aimed at elucidating the role played by those factors may help develop more effective RNA editing therapeutics.
Subcellular localization of ADARs
The editing activity of ADARs is modulated not only by their expression profiles but also by their subcellular localization, which critically influences substrate accessibility. ADAR1p110 and ADAR2, which are predominantly nuclear, have been reported to constantly shuttle in and out of the nucleolus, where they bind but do not edit ribosomal RNA.54–56 Upon expression of editing-competent GRIA2 in cells, endogenous ADAR1 and ADAR2 are redistributed from the nucleolus to sites of substrate transcript accumulation. 55 Similarly, increased translocation of endogenous ADAR2 from the nucleolus to the nucleoplasm has been shown to lead to elevated editing of its substrates. 56 These findings collectively suggest that the nucleolar localization of the ADARs may serve to sequester the enzymes, thereby modulating their intracellular levels and enzymatic activity toward their substrates in the nucleoplasm.56,57 Alternatively, microRNAs (miRNAs), some of which are present in the nucleolus, can undergo A-to-I editing in their precursor form, and the nucleolar localization of the ADARs may possibly be linked to their activity on these RNAs.58–61 Unlike ADAR1p110 and ADAR2, ADAR1p150 is predominantly cytoplasmic, enabling it to access and edit cytosolic dsRNA substrates, although it is also detectable in the nucleus. 35 Reflecting the difference in subcellular localization between the two ADAR1 isoforms, Kleinova et al. identified isoform-specific editing patterns in their study of editing-deficient mouse cells transfected with individual ADAR1 isoforms, which they attributed primarily to the distinct intracellular localization of the isoforms. 62 They also reported that Zα—the domain that mainly distinguishes ADAR1p150 structurally from ADAR1p110—only minimally contributes to its editing specificity, 62 supporting the role of intracellular ADAR localization in determining editing specificity. ADAR3 is primarily nuclear, with notable nucleolar localization observed during neuronal differentiation. 63 This nuclear localization is facilitated by an R-rich motif (R-domain), which binds to single-stranded RNA (ssRNA) and acts as a functional NLS by interacting with the importin karyopherin subunit alpha 2 (KPNA2).17,64 Interestingly, the main splice form of ADAR2 shows no interaction with KPNA2 but does interact with other importin alpha family members such as KPNA1 and KPNA3. 64 In contrast, a minor splice variant of ADAR2, ADAR2R, displays an importin alpha interaction profile indistinguishable from that of ADAR3.64,65 These findings suggest that the expression of specific importin alpha proteins may lead to the nuclear import of certain ADAR proteins or splice variants, influencing the resulting RNA editing landscape. 64 The in vivo regulation of RNA editing may thus depend at least partially on the coordinated importin alpha protein and ADAR isoform expression. 64
ADAR sequence preference and structural selectivity
ADAR editing occurs in both coding and noncoding regions of RNA, which can lead to codon changes (recoding), alterations of splice site selection, and modification of miRNA-binding sites.43,58 Most RNA editing occurs in noncoding regions, with over 90% of the editing sites in the human transcriptome located within Alu repetitive elements, which are the primary target of ADAR1.66,67 In HEK293 cells, it has been shown that Alu elements edited by ADAR1 are mainly found in introns (48%) and 3′ untranslated regions (UTRs; 37%), whereas less than 0.1% of ADAR1 edits occur in open reading frames. 67 The two ADAR1 isoforms exhibit distinct regional specificities: ADAR1p110 binds and edits intronic regions, whereas ADAR1p150 shows a preference for exonic sequences and 3′ UTRs. 62 The editing activity of ADAR1 extends beyond Alu repeats to other transposable elements, including endogenous retroviruses (ERVs) and long interspersed nuclear elements (LINEs) such as L1s. 67 These repetitive sequences along with Alu elements are frequently arranged in inverted orientations, and when two such elements are located in close proximity, their transcription leads to the formation of secondary RNA structures such as hairpins, which are the major substrates for ADAR1. 68 The functional implications of Alu editing have been reviewed elsewhere, but include—but are not limited to—exonization of intronic Alu sequences, retention of edited Alu dsRNAs in paraspeckles, suppression of the IFN response, and heterochromatin formation and gene silencing. 58 Recently, Katrekar et al. showed that Alu repeats can be used as endogenous ADAR-recruiting domains within gRNAs, 69 demonstrating their applicability in programmable RNA editing. ADAR2, unlike ADAR1, primarily functions in editing nonrepetitive coding sites, particularly in the brain. 53 Two transcripts known to contain ADAR2-specific editing sites include cytoplasmic FMR1 interacting protein 2 (CYFIP2) and GRIA2. Editing of a single site in the CYFIP2 coding sequence leads to a lysine (K) to glutamic acid (E) substitution, which predominantly occurs in the brain, where it influences neuronal axon and spine maturation.70,71 The GRIA2 transcript is edited at the Q/R site exclusively by ADAR2, which is well established. 72 This site is known to be edited to nearly 100% in the mammalian brain, and a reduction in its editing leads to increased calcium permeability of the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor, seizures, and early lethality in mice, indicating the critical importance of Q/R site editing in normal brain function.73,74 Another well-studied site in GRIA2 targeted by ADAR2 is the R/G site, where RNA editing results in a codon change from R to glycine (G). This site, which is also edited by ADAR1, holds particular significance owing to its utility as a physiological recruitment sequence for endogenous or full-length exogenous ADAR in gRNA designs.14,45,75,76
While ADARs are dependent upon dsRNA secondary structures for their substrate recognition, they preferentially select certain adenosines for modification based on the local sequence context. 77 For both ADAR1 and ADAR2, the nucleotide immediately upstream (5′) of the target adenosine exerts the greatest influence, although adjacent bases also contribute to editing site selection. 78 The two enzymes share the same 5′ neighbor preference of U > A > C > G, but their 3′ neighbor preferences differ slightly: ADAR1 favors G > C ≈ A > U, while ADAR2’s preference follows the order G > C > U ≈ A. 78 As a result of these preferences, motifs containing a 5′ guanosine, such as 5′-GAN-3′, are poorly edited, whereas the 5′-TAG-3′ (stop codon) motif is highly favored for editing. 14 It should be noted that, however, beyond the neighboring sequence, the base-pairing status of the target adenosine is an important determinant of editing efficiency. 70 For instance, Wong et al. reported that both ADAR1 and ADAR2 preferentially edited adenosine paired with cytidine (A:C mismatch) over that paired with adenosine (A:A), guanosine (A:G), or uridine (A:U). 79 Taking advantage of these characteristics, ADAR-based systems have been employing an A:C mismatch at the intended editing site to enhance target specificity and an A:G mismatch opposing nontarget adenosines to reduce or avoid bystander editing.7,8,14,69,80–83
ADAR editing efficiency is also affected by the length of RNA substrates. It has been shown that ADAR exhibits editing activity on inter- and intramolecular dsRNA longer than 20 bp. 84 However, nonselective A-to-I editing has been observed in long (>100 bp), perfectly matched dsRNAs, with an editing rate of 50%–60%, whereas shorter dsRNAs are edited more selectively at a rate below 10%.77,85 The nonspecific editing of long dsRNAs may stem from the sequence-independent binding properties of dsRBDs within ADAR proteins. 86 Therefore, incorporating secondary structures such as internal loops into gRNAs, particularly those with long antisense domains, may represent an effective strategy to reduce bystander editing, as previously reported.8,69 The target adenosine identified is extruded from the dsRNA helix into the ADAR active site within the DD through a base-flipping mechanism. 26 Site-specific RNA editing in vivo can be achieved by the DD independent of dsRBDs, and DDs from both ADAR1 and ADAR2 have been used in exogenous ADAR systems.14,80,87,88 In addition, the ADAR mutants ADAR1 (E1008Q) and ADAR2 (E488Q), which carry hyperactive E to Q substitutions in the DD, are widely used to enhance editing efficiency due to their increased base-flipping activity.89,90
RNA Editing with Exogenous ADARs
CRISPR-based/-inspired ADAR systems
The CRISPR–Cas13 system represents a transformative advancement in RNA editing, with significant implications for RNA therapeutics. As a class 2 type VI system, it is distinguished from other CRISPR–Cas subtypes by its ability to target RNA for editing without altering DNA sequences, utilizing a single effector protein, Cas13, guided by gRNA. 91 Among other Cas13 effectors, Cas13b (subtype VI-B) and Cas13d (subtype VI-D) have been reported to exhibit high specificity, making them attractive tools for RNA editing applications.80,92 Since the report of the first CRISPR-based RNA editing tool, in which a catalytically inactive Cas13b (dCas13b) was fused to ADAR, 80 different ADAR-based programmable RNA editing platforms leveraging CRISPR–Cas13 (Table 1) have been developed, offering a promising avenue for correcting disease-causing mutations without permanent genomic alteration.
Comparison of the Main Features of Adenosine Deaminase Acting on RNA-Based RNA Editing Systems
The RESCUE systems support both A-to-I and C-to-U RNA editing; however, the editing efficiency and off-target data presented pertain to A-to-I editing only.
AAV, adeno-associated virus; ABI, abscisic acid insensitive; ADAR, adenosine deaminase acting on RNA; ADARDD, adenosine deaminase acting on RNA deaminase domain; adRNA, ADAR-recruiting RNA; AIMers, A-to-I RNA base editing oligonucleotides; ASO, antisense oligonucleotide; BG, O6-benzylguanine; CA, 1-chloroalkane; cadRNAs, circular ADAR-recruiting guide RNAs; CIRTS, CRISPR–Cas-inspired RNA targeting system; CLUSTER, clustered ADAR-recruiting guide RNAs for efficient RNA editing; CRISPR, clustered regularly interspaced short palindromic repeats; ecDHFR-DD, Escherichia coli dihydrofolate reductase destabilization domain; ecRESCUE, ecDHFR-regulated RESCUE; eRESCUE, enhanced RESCUE; GalNAc, N-acetylgalactosamine; gRNA, guide ribonucleic acid; LEAPER, leveraging endogenous ADAR for programmable editing of RNA; LNP, lipid nanoparticles; MIRROR, mimicking inverted repeats to recruit ADARs using engineered oligoribonucleotides; NES, nuclear export signal; NLS, nuclear localization signal; N3U, N-3-uridine; PN, nitrogen-containing, phosphoryl guanidine-based; PS, phosphorothioate; PYL, pyrabactin resistance-like; REPAIR, RNA editing for programmable A to I replacement; RESCUE, RNA editing for specific C-to-U exchange; RESTORE, recruiting endogenous ADAR to specific transcripts for oligonucleotide-mediated RNA editing; R/G, arginine- and glycine-rich; scAAV, self-complementary adeno-associated virus; ssRNA, single-stranded RNA; TMP, trimethoprim.
RNA Editing for Programmable A to I Replacement
The RNA Editing for Programmable A to I Replacement (REPAIR) system is the first programmable RNA editing platform based on CRISPR technology, comprising dCas13b from Prevotella sp. P5-125 (dPspCas13b) fused to the ADAR deaminase domain (ADARDD), as shown in Figure 3A. 80 Guided by a CRISPR RNA containing a mismatched cytidine opposite the target adenosine, the dPspCas13b protein fused to a hyperactive ADARDD—specifically the ADAR2DD carrying the E488Q mutation—has been shown to enable precise A to I editing. 80 This system, designated as REPAIR version 1 (REPAIRv1), corrected two disease-relevant mutations in vitro: W293X in the arginine vasopressin receptor 2 (AVPR2) gene, associated with X-linked nephrogenic diabetes insipidus, at 35% efficiency, and W506X in the Fanconi anemia complementation group C (FANCC) gene, associated with Fanconi anemia, at 23% efficiency. 80 In addition, REPAIRv1 achieved up to 28% correction of other disease-relevant G>A mutations at 33 sites, 80 demonstrating its successful and therapeutically relevant application in mammalian cells. However, due to its substantial off-target activity, the system was subsequently engineered with two mutations (T375G and E488Q) in ADARDD to enhance specificity. 80 This advanced version termed REPAIRv2, exhibited more than 919-fold higher specificity than REPAIRv1 by significantly reducing transcriptome-wide off-target effects. 80 The increase in specificity, however, came at the expense of on-target editing efficiency as the T375G mutation that differentiates REPAIRv2 from REPAIRv1 weakens the ADARDD. 93 To address this trade-off, Liu et al. developed REPAIRx (Vx), which integrates the strengths of the two earlier versions—the robust activity of REPAIRv1 and the high specificity of REPAIRv2. 93 REPAIRx is based on CasRx, a Cas13d family protein, and is characterized by its nuclear localization and the insertion of ADARDD with the E488Q mutation into the middle of dCasRx. 93 This optimized design enables REPAIRx to achieve precise and highly specific A-to-I editing, surpassing previous versions in performance on both mRNA and nuclear RNA targets and significantly expanding the RNA editing toolkit. 93
Exogenous ADAR-based RNA editing platforms.
RNA Editing for Specific C-to-U Exchange
The RNA Editing for Specific C-to-U Exchange (RESCUE) system, initially developed for C-to-U RNA editing, utilizes a dCas13b from Riemerella anatipestifer (dRanCas13b) fused to an engineered ADAR2DD, which has been modified via serial mutagenesis to enable cytidine deamination (Fig. 3B). 7 RESCUE contains 16 mutations distributed throughout the structure of ADAR2DD, resulting in both direct effects on RNA binding within the catalytic pocket and indirect effects. 7 In particular, its activity is critically influenced by mutations in the catalytic core (V351G, K350I) and in the region contacting the RNA target (S486A, S495N) and is maintained—or even enhanced—with C-terminal truncations of dRanCas13b, enabling its minimization for adeno-associated virus (AAV) packaging and delivery. 7 The RESCUE system has demonstrated up to 42% editing efficiency in mammalian cells, with further improvement accomplished through the development of a variant system, termed eRESCUE, by Li et al.7,94 The eRESCUE system employs dPspCas13b in place of dRanCas13b and exhibits more efficient A-to-I and C-to-U RNA base editing than the RESCUE editor. 94 The mutations in RESCUE enable the catalytic pocket to accommodate adenosine or cytidine, allowing RESCUE to catalyze both A-to-I and C-to-U RNA editing. 7 Although this multiplexed editing capacity offers a significant functional advantage, it is associated with an elevated frequency of off-target events, with A-to-I off-target edits occurring at a level comparable to that for REPAIRv1 and C-to-U off-target edits observed to a lesser extent. 7 A transcriptome sequencing study of HEK 293T cells revealed that RESCUE induces many A-to-I and C-to-U off-target SNVs that affect mRNA, circular RNA, and lncRNA expression, as well as their interacting networks. 95 To improve the specificity of RESCUE, ADAR2DD was further modified through rational mutagenesis at residues interacting with the RNA target. 7 The resulting mutant with the highest specificity, S375A,—referred to as RESCUE-S—showed approximately 12-fold and 1.8-fold reductions in A-to-I and C-to-U off-target edits, respectively, while maintaining comparable efficiency to RESCUE.1,7 The off-target editing frequency was further reduced in the ecRESCUE system, in which the Escherichia coli dihydrofolate reductase destabilization domain (ecDHFR-DD) is fused to the C-terminus of RESCUE. 96 This fusion enables inducible control of editing activity via the small molecule trimethoprim (TMP), a DHFR inhibitor that stabilizes the otherwise unstable ecDHFR-DD. 96 Consequently, the expression of dRanCas13b and ADAR2DD is TMP-dependent, allowing temporal regulation of RNA editing. 96 This inducible system enables dual base conversion (A-to-I and C-to-U), while significantly reducing the incidence of off-target single-nucleotide polymorphisms without compromising editing efficiency, making it a promising approach for achieving improved and controllable RNA editing outcomes. 96
CRISPR–Cas-inspired RNA targeting system
Drawing inspiration from Cas13 proteins and aiming for clinical applicability, Rauch et al. developed a modular and programmable RNA targeting platform known as the CRISPR–Cas-inspired RNA targeting system (CIRTS).97,98 CIRTS, which is constructed entirely from human protein parts, consists primarily of two components: a gRNA and a tripartite protein comprising a ssRNA-binding region fused to a hairpin RNA-binding region, which is further fused to an effector domain (Fig. 3C).97,98 The gRNA contains a sequence complementary to the target RNA as well as a structural element that interacts with the hairpin RNA-binding region, while the ssRNA-binding region interacts with the displayed gRNA to stabilize and protect it prior to target association. 97 Serving as the core of the system, the hairpin RNA-binding region exhibits high-affinity, selective binding to a specific RNA structure displayed on the engineered gRNA. 97 By facilitating target RNA binding, the displayed guide sequence brings the effector domain into close proximity, allowing it to act on the target sequence.97,99 Modular substitution of effector domains enables the system to be adapted for diverse applications such as RNA degradation, translational regulation, and base editing, highlighting its versatility. 97 For RNA editing, CIRTS was employed in a reporter assay to deliver the catalytic domain of human ADAR2 (wild-type and the E488Q mutant) to a G-to-A mutation site in the coding region of firefly luciferase, which causes a premature stop of translation and abolishes measurable luciferase activity. 97 This targeting led to gRNA-dependent restoration of luciferase activity with both wild-type and ADAR2 (E488Q), 97 providing functional evidence for CIRTS as a platform for mediating site-specific RNA editing. Moreover, the small size of CIRTS—smaller than many comparable CRISPR-based systems—allows efficient packaging into in vivo delivery vehicles such as AAVs, while its human-derived nature alleviates immunogenicity concerns, enhancing its potential for therapeutic applications. Building upon these advantages, CIRTS was further engineered to enable temporal regulation through a small molecule-dependent dimerization system, known as the abscisic acid (ABA) system, which relies on rapid heterodimerization between abscisic acid insensitive (ABI) and pyrabactin resistance-like (PYL) domains upon ABA addition.98,100 In this ABA-inducible system, termed the CIRTS biosensor, the targeting component of CIRTS (a ssRNA-binding protein and hairpin-binding protein) and the effector domain are fused to ABI and PYL, respectively. 100 The targeting component binds to a gRNA and engages the target RNA, after which ABA induces ABI–PYL heterodimerization to recruit the effector domain to the target site and elicit RNA modulation. 100 The CIRTS biosensor incorporating CIRTS-ABI and PYL-hADAR2 (E488Q) successfully edited endogenous transcripts such as glyceraldehyde-3-phosphate dehydrogenase and peptidylprolyl isomerase B in an ABA-dependent manner. 100 It also reversed known disease-causing mutations in vitro, including the R106Q mutation in methyl-CpG-binding protein 2 (MECP2), associated with Rett syndrome, and the W1262X mutation in APC regulator of Wnt signaling pathway (APC), associated with familial adenomatous polyposis. 100 Importantly, the system’s ABA-dependent RNA base editing activity was validated in a murine model, demonstrating its functionality in vivo. 100 However, low-to-moderate editing efficiency on endogenous transcripts remains a limitation, warranting additional optimization for therapeutic use.
Non-CRISPR-based ADAR systems
λN-BoxB-ADAR
One of the early approaches for SDRE exploits the high-affinity interaction between the λN protein from bacteriophage lambda and the BoxB RNA hairpin structure. 101 To harness this interaction for targeted editing, the λN peptide, which is the 22-amino-acid RNA-binding domain of the λN protein, was fused to ADAR2DD, while a gRNA was designed with BoxB hairpins incorporated within an antisense guide sequence containing a mismatched C opposite the target A.101,102 Binding of the λN–ADAR2DD fusion protein to the BoxB-gRNA directs ADAR2’s deaminase activity to the editing site (Fig. 3D). 102 This system offers a significant advantage owing to the small size of the editing complex, enabling its packaging with several copies of the gRNA into a single AAV.1,82 However, it exhibits substantial off-target effects, particularly with 4λN-ADAR2DD (E488Q), which features four λN peptides. 103 In their study of the transcriptome-wide effects of various combinations of editing enzymes and gRNAs, the Rosenthal group reported that the presence of the E488Q mutation was generally associated with a considerable increase in off-target editing, with enzymes carrying this mutation producing many off-target events even in the absence of a gRNA. 103 This effect, which may be attributed to editase overexpression, can be mitigated by introducing an NLS to the enzyme and by lowering the gRNA concentration.1,103 Despite its off-target effects, the system can achieve editing efficiencies of up to 72% when optimized with four λN peptides and two BoxB hairpins in the gRNA, with demonstrated efficacy in both cultured cells and in mouse models.82,104–106 In addition, given its successful application in correcting disease-relevant mutations,82,101,105,106 the λN-BoxB-ADAR approach represents a potential avenue for therapeutic development, albeit with limitations that must be addressed.
SNAP-ADAR and HALO-ADAR
The SNAP-ADAR platform achieves site-directed A-to-I RNA editing through fusion of the ADARDD with a self-labeling SNAP-tag, which covalently binds to O6-benzylguanine (BG)-conjugated to synthetic gRNAs (Fig. 3E).107,108 Chemically modified for nuclease resistance, these gRNAs hybridize to target transcripts via Watson–Crick base pairing and enable targeted recruitment of ADAR to a specific adenosine.107,108 The use of gRNAs for directing deaminase activity, in lieu of ADAR’s native dsRNA-binding domains, allows for improved specificity and programmability. One notable advantage of SNAP-ADAR is reduced off-target editing within the gRNA/mRNA duplex via gRNA chemical modifications. 87 Off-target activity, observed only for adenosine-rich triplets (AAC, AAA, UAA, CAA)—particularly CAA and mainly with SNAP-ADAR2—was reduced at all sites through chemical modifications such as 2′-methoxy and 2′-fluoro (2’-F), with a greater reduction for the CAA triplet, and without affecting on-target editing. 87 Another advantage of this system is its high editing efficiency, with levels reaching up to 90% when a hyperactive mutant version of ADAR2 is employed. 87 Leveraging these advantageous properties, SNAP-ADAR has been applied not only in mammalian cell culture but also in embryos. Intriguingly, Hanswillemenke et al. developed a strategy to induce the covalent assembly of the gRNA and ADAR in a light-dependent manner, thereby enabling photocontrol of SDRE. 109 This strategy has been successfully implemented in HEK 293T cells and developing embryos of the annelid Platynereis dumerilii with photoactivation of RNA editing achieved in a dose-dependent manner, particularly in vitro, 109 thereby demonstrating its potential to enable fine-tuned regulation of RNA editing. Besides the SNAP-tag, another self-labeling tag, HALO, has been employed to allow for orthogonal RNA editing, requiring gRNAs to be modified with a 1-chloroalkane (CA) moiety (Fig. 3E). 110 Stroppel et al. generated bifunctional gRNAs carrying both the BG and the CA moieties capable of co-recruiting two editing effectors (ADAR1 and ADAR2), which resulted in optimal editing efficiency with an extended substrate scope and moderate global off-target effects. 110 This flexibility in the linker chemistry, along with the small size of self-labeling proteins, 110 makes this system modular and suited for multiplexed or delivery-constrained RNA editing applications. However, gRNAs with chemical moieties are genetically nonencodable,1,111 which must be carefully considered when evaluating the therapeutic potential of SNAP- and HALO-ADAR-based RNA editing systems.
RNA Editing with Endogenous ADARs
Exogenous ADAR-based editing platforms present several challenges, including aberrant effector activity, delivery constraints, and immunogenic risk. 1 Recent advancements in RNA editing technologies have made the use of endogenous ADAR enzymes for targeted RNA editing possible by employing gRNAs that hybridize to the target transcript and form mismatches that are recognized and edited by these enzymes (Table 1). 4 Harnessing native ADAR activity enables high specificity and reduced off-target effects, 4 presenting a safer and more efficient alternative to exogenous ADAR delivery.
Leveraging Endogenous ADAR for Programmable Editing of RNA
The Leveraging Endogenous ADAR for Programmable Editing of RNA (LEAPER) system harnesses engineered linear ADAR-recruiting RNAs (arRNAs) to direct native ADAR1 or ADAR2 enzymes to specific adenosines within target RNAs (Fig. 4A). 8 The length of the arRNA, which typically ranges from 71 to 200 nucleotides, positively correlates with editing efficiency, with a minimum of 71 nucleotides required and peak efficiency observed within the 111–191 nucleotide range.4,8,13 A single A–C mismatch is introduced at the intended editing site within the arRNA to enhance editing specificity, with delivery achievable through plasmids, viral vectors, or as synthetic oligonucleotides. 8 This approach was shown to be effective across various human primary cells and cell lines at efficiencies of up to 80%, with rare global off-targets.8,13 The Wei group reported an approximately 3.1-fold increase in editing efficiency with LEAPER 2.0, an updated version of LEAPER that employs circular arRNAs (circ-arRNAs) for greater stability (Fig. 4A). 81 Notably, LEAPER 2.0 enabled RNA editing to persist for up to 21 days in vitro and achieved sustained and enhanced editing at the target site when delivered via AAV or self-complementary AAV (scAAV). 81 Furthermore, bystander off-target adenosine editing was almost completely abolished by eliminating pairings of uridines with off-target adenosines, 81 indicating a marked improvement in editing specificity. Importantly, LEAPER 2.0 has also demonstrated efficacy in correcting a disease-causing mutation in vivo, positioning it as a precise and efficient A-to-I editing platform with promising therapeutic potential.
Endogenous ADAR-based RNA editing platforms.
Recruiting Endogenous ADAR to Specific Transcripts for Oligonucleotide-mediated RNA Editing
Extending their work on the SNAP-ADAR system, the Stafforst group developed an innovative platform, Recruiting Endogenous ADAR to Specific Transcripts for Oligonucleotide-mediated RNA Editing (RESTORE). 45 This system enables programmable, site-specific RNA editing by recruiting endogenous ADAR enzymes via chemically stabilized antisense oligonucleotides (ASOs). 45 RESTORE utilizes an ASO composed of two domains: an ADAR-recruiting domain (R/G motif) and a specificity domain containing chemical modifications such as 2′-O-methylations (2’-O-Me), phosphorothioate (PS), and locked nucleic acid, as well as a modification gap opposite the editing site (Fig. 4B). 45 It was applied to standard human cell lines and primary cells, achieving editing efficiencies ranging from 4% to 63%, which were further increased by IFN-α treatment to upregulate ADAR1p150, while maintaining minimal off-target effects.13,45 Furthermore, the platform was successfully employed to correct disease-relevant mutations in vitro. 45
Building on this original platform (retroactively referred to as “RESTORE 1.0”), the Stafforst group developed an improved system, RESTORE 2.0, which addresses key limitations of the earlier design and enhances the translational potential of oligonucleotide-mediated RNA editing. 112 Unlike RESTORE 1.0, which relied on longer, partially chemically modified ASOs with low metabolic stability, RESTORE 2.0 employs shorter (30–60 nucleotides), fully chemically stabilized ASOs devoid of the ADAR-recruiting domain, incorporating commercially available and clinically used modifications such as phosphate (PO)/PS, 2′-O-Me, 2′-H, and 2′-F, and eliminates the need for synthetically demanding stereopure linkages. 112 Systematic optimization of chemical modification patterns enhanced nuclease stability while preserving hybridization specificity and minimizing global off-target editing. 112 The refined ASOs exhibited robust editing activity in cell lines, primary cells, and an animal model, achieving up to 80% and 25.8 ± 6.2% editing efficiencies in vitro and in vivo, respectively. 112 Importantly, they have been shown to recruit endogenous ADARs to edit a disease-causing mutation both in vitro and in vivo, highlighting their therapeutic potential. 112 Although RESTORE ASOs are genetically nonencodable, they provide a broad sequence and chemical space to refine their pharmacological properties, making them a promising platform for therapeutic RNA editing.
Clustered ADAR-recruiting guide RNAs for efficient RNA editing
To address the limitations of the two earlier endogenous ADAR-based systems—RESTORE (lack of genetic encodability, moderate editing efficiency, and IFN-α–induced ADAR1p150 expression dependence of editing yields) and LEAPER (substantial bystander off-target editing)—the Stafforst group devised an alternative strategy termed clustered ADAR-recruiting guide RNAs for efficient RNA editing (CLUSTER). 75 This approach, building on prior R/G-gRNA designs, incorporates a cluster of single-stranded RS, which bind to the target mRNA in various regions, distal to the intended editing site and each other (Fig. 4C).4,75 By leveraging the cooperative interaction between the RS and the specificity domain, the design facilitates rapid and strong guide-target RNA binding while supporting highly flexible sequence selection for gRNAs to optimize their properties. 75 Bystander editing is avoided by choosing RS binding sites that minimize editable adenosine bases in the gRNA–target RNA duplex.75,113 As a result of these optimizations, CLUSTER has been reported to achieve up to 45% editing efficiency in vitro without bystander editing, while in vivo gRNA delivery via hydrodynamic tail vein injection yielded up to 10% editing of reporter constructs in the mouse liver. 75 Although no bystander editing was observed under the reported in vitro conditions, the system was further optimized to address potential off-target effects by exploiting the sequence context dominantly associated with bystander editing, 5′-UAN triplets, into which G•U wobble base pairs are strategically introduced. 113 This approach, combined with a circularized CLUSTER format (Fig. 4C), enables highly efficient and precise editing of a disease-relevant loss-of-function mutation in the MECP2 transcript, achieving up to 87% editing in vitro. 113 Moreover, virus-mediated gRNA delivery leads to functional rescue of MECP2 in a murine disease model, with editing yields of up to 19% and excellent control of bystander editing. 113 Beyond its precision and efficiency, the system offers versatility, making it broadly applicable to existing A-to-I RNA base editing platforms and complement other suppression approaches, such as G•A mismatches and uridine (U) depletion. 113
Circular ADAR-recruiting guide RNAs
Circular ADAR-recruiting guide RNAs (cadRNAs) represent another important innovation in RNA editing technology. Prior to their development, the Mali group demonstrated that simple long antisense RNAs greater than 60 nucleotides in length are capable of recruiting endogenous ADARs. 83 These ADAR-recruiting RNAs (adRNAs) have, however, shown limited editing efficiency, partly due to their short half-life and corresponding target residence times. 69 To overcome these limitations, they engineered cadRNAs using a technique described by Litke et al., flanking linear adRNAs with twister ribozymes that autocatalytically cleave and produce termini ligated by the endogenous RNA ligase RtcB, resulting in circularized gRNAs (Fig. 4D).69,114 The enhanced stability and robust expression of cadRNAs lead to vastly improved efficiency and durability of RNA editing at multiple sites in both untranslated and coding regions across different cell lines, with high transcriptome-wide specificity. 69 In addition, by incorporation of interspersed loops in the antisense domains, greater transcript-level specificity for the target adenosine was achieved, accompanied by decreased bystander editing. 69 Importantly, the enhanced performance of cadRNAs extended beyond in vitro systems, yielding robust, persistent, and highly transcript-specific editing of a disease-relevant mutation in a mouse disease model. This in vivo efficacy, combined with their genetic encodability and compatibility with various delivery modalities, underscores their therapeutic potential.4,69 Notwithstanding their promise, cadRNAs require further refinement to fully achieve their therapeutic potential, including enhancing their editing efficiency through the incorporation of additional ADAR recruitment domains, optimizing their design based on target-specific sequence and structural contexts, and evaluating long-term immune responses in vivo.
A-to-I RNA base editing oligonucleotides
A-to-I RNA base editing oligonucleotides (AIMers), developed by Vargeese and colleagues, are distinguished from other ADAR-based RNA editors by their extensive chemical modifications. These short oligonucleotides feature fully modified chimeric backbones containing stereopure PS and nitrogen-containing phosphoryl guanidine (PN) linkages (Fig. 4E). 115 Efficient and specific A-to-I editing of endogenous transcripts is achieved through this design via recruitment of endogenous ADAR enzymes, including the constitutively and ubiquitously expressed ADAR1p110 isoform. 115 In vitro, they have been shown to enhance potency and editing efficiency 100-fold compared with uniformly PS-modified counterparts. 115 In line with these results, up to 50% editing was achieved in nonhuman primate liver using N-acetylgalactosamine (GalNAc)-modified AIMers in vivo, with no bystander editing observed and sustained activity for at least 1 month. 115 Unlike the observations made with RESTORE, another chemically modified platform relying on endogenous ADARs, editing by a stereopure AIMer targeting β-actin (ACTB) was not significantly affected by IFN-α treatment. 115 Furthermore, knockdown of both ADAR1 isoforms or of p150 alone significantly reduced editing, while depletion of ADAR2 had no effect, demonstrating that AIMer-mediated editing is substantially supported by ADAR1p110 in vitro. 115 The potency and efficiency of AIMers were further improved by new designs integrating sugar, backbone, and base modifications. 116 Substitution of cytidine (C) with N-3-uridine (N3U) at the “orphan base” position opposite the edit site had a particularly significant impact on RNA editing efficiency, with AIMers featuring N3U combined with novel 2′ sugar chemistry and backbone modifications achieving maximal editing of over 85% in mouse hepatocytes in vitro and approximately 70% in the mouse liver in vivo. 116 The effect of N3U on RNA editing was demonstrated across multiple targets with different nearest neighbor sequences, indicating its ability to at least partially overcome the inherent sequence bias of ADAR enzymes. 116 Despite variation in its effect depending on AIMer and target chemical and sequence contexts, the versatile nature of N3U offers useful flexibility for AIMer development, expanding opportunities to optimize critical pharmacological properties without compromising potency. 116
Mimicking inverted repeats to recruit ADARs using engineered oligoribonucleotides
Mimicking inverted repeats to recruit ADARs using engineered oligoribonucleotides (MIRROR) constitutes an innovative approach that enhances endogenous ADAR recruitment by emulating optimal RNA structures found in naturally highly edited inverted repeats. 117 MIRROR gRNAs are designed to form intermolecular structures with target RNAs that closely resemble those of inverted Alu repeats, featuring a central mimic region surrounding the editing site and binding regions at the 5′ and/or 3′ ends to improve target affinity, with adjustable lengths for optimized performance (Fig. 4F). 117 By structurally mimicking ADAR’s natural RNA substrates, these gRNAs have been shown to enhance ADAR recruitment and editing, achieving up to a 5.7-fold increase in editing efficiency in vitro in multiple human cell types. 117 Furthermore, MIRROR gRNAs encapsulated in lipid nanoparticles (LNPs) demonstrated robust, dose-dependent in vivo editing in mice by endogenously expressed ADARs, yielding efficiencies of 73%–81% that were sustained for at least seven days. 117 Beyond its demonstrated in vitro and in vivo efficacies, MIRROR is notable for its versatility, supporting both short, chemically synthesized gRNAs with chemical modifications and long, biologically generated gRNAs, outperforming existing RNA editing platforms in both forms. 117
Disease-Relevant Applications of ADAR-Based RNA Editing Platforms
ADAR-based RNA base editors have opened new avenues for therapeutic intervention across a wide range of diseases by enabling precise, programmable, and transient RNA modifications. Advances in chemical design and delivery technologies have further expanded their applicability to clinically relevant targets. This section offers a comprehensive overview of key disease-focused applications that highlight the increasing therapeutic potential of ADAR-mediated RNA editing (Table 2).
Adenosine Deaminase Acting on RNA-Based RNA Editing Platforms Used in Disease Models
ceRBE, compact and efficient RNA base editor; CFTR, CF transmembrane conductance regulator; circ-arRNA, circular ADAR-recruiting RNAs; Dmd, Dystrophin; eGFP, enhanced green fluorescent protein; IDUA, alpha-L-iduronidase; MCP, MS2 coat protein; Otc, ornithine transcarbamylase; PINK1, PTEN induced kinase 1; RFU, relative fluorescence units; RNA, ribonucleic acid; SERPINA1, serpin family A member 1.
Alpha-1 antitrypsin deficiency
Alpha-1 antitrypsin deficiency (AATD) is a genetic disorder most frequently caused by mutations in serpin family A member 1 (SERPINA1), which reduce serum levels of alpha-1 antitrypsin—whose main function is to protect the lung from proteolytic damage by neutrophil elastase—and cause lung and liver disease. 118 The Stafforst group used the RESTORE 1.0 platform to edit the PiZZ mutation (E342K) in SERPINA1, the most common cause of AATD, achieving in vitro editing yields of 18%–29% in cell lines and elevated secretion of alpha-1 antitrypsin (AAT). 45 Extending this work, the group employed the improved RESTORE 2.0 system, which yielded editing efficiencies of 40%–50% and 25.8% ± 6.2% in mutant SERPINA-1expressing cell and mouse models, respectively, accompanied by restored AAT levels. 112 Similarly, another endogenous ADAR-based system, AIMers, demonstrated efficient in vitro editing, as shown in a study by Monian et al. 115 In this study, the use of modified AIMers in primary hepatocytes from NSG-PiZ mice expressing the human SERPINA1 E342K mutation accomplished editing rates of 68%–75% and led to increased AAT secretion. 115 In addition, a recent report demonstrated effective RNA editing by MIRROR gRNAs in primary hepatocytes from an AATD mouse model, 117 further highlighting the potential of RNA editing to correct SERPINA1 mutations.
Cancer
RNA editing is particularly advantageous in cancer therapy because it enables selective correction or modulation of oncogenic transcripts within genetically heterogeneous tumors, offering a reversible and transient approach that avoids permanent genomic alterations—an important consideration given the high mutational burden, clonal diversity, and genomic instability characteristic of many cancers. 119 Over 50% of human cancers harbor mutations in the tumor protein p53 (TP53) gene, which encodes the critical tumor suppressor protein p53, including the clinically relevant W53X (c.158G>A) nonsense mutation.120,121 LEAPER was shown to be effective in correcting the W53X mutation in HEK293T TP53−/− cells cotransfected with reporter and mutant TP53-expressing plasmids with an editing efficiency of approximately 25%–35% and restoring TP53 transcriptional activity and full-length p53 protein production. 8 Notably, LEAPER 2.0 enhanced the efficiency for editing the same mutation to 30%–50%, indicating that circularization of ADAR-recruiting RNAs positively contributes to improved editing performance in cancer contexts, as also demonstrated in a recent study in which LNP-encapsulated circular ADAR-recruiting RNA achieved W53X mutation correction efficiencies of over 73% and 48% in triple-negative breast cancer cells and tumor-bearing mouse models, respectively.81,122 This study also confirmed restoration of full-length p53 protein expression and its functional activity, as well as improved sensitivity to paclitaxel chemotherapy following RNA editing in vivo, supporting the potential exploitation of circular ADAR-recruiting RNAs for the treatment of TP53 mutant cancers. 122
Cystic fibrosis
Cystic fibrosis (CF) is a genetic disorder caused by mutations in the CF transmembrane conductance regulator (CFTR) gene that result in defective chloride ion transport across epithelial membranes, thickened mucus, chronic lung infections, pancreatic dysfunction, and reduced life expectancy. 123 RNA editing enables targeted correction of CFTR mRNA transcripts, offering a transient and precise strategy to restore protein function, particularly in cases involving rare or nonresponsive mutations not addressed by current CFTR modulators. Preclinical studies using ADAR-based RNA editing tools have demonstrated the feasibility of correcting CFTR nonsense mutations, highlighting their promise as next-generation therapeutics. Specifically, the λN-BoxB-ADAR system achieved approximately 20% correction of the W496X nonsense mutation, with restoration of full-length CFTR protein, recovery of functional chloride currents, and no evidence of off-target editing in Xenopus oocytes. 101 Melfi et al. have demonstrated modest efficacy of REPAIRv2 in correcting the UGA premature stop codon in Fischer rat thyroid cells engineered to express CFTRW1282X, as well as in human IB3-1 airway epithelial cells harboring a naturally occurring compound heterozygous mutation (F508del/W1282X) that do not express detectable endogenous CFTR protein, with evidence suggesting partial restoration of the CFTR protein. 124 Similarly, the endogenous ADAR-based system RESTORE was effective in correcting the CFTR nonsense G542X (UGG > UGA) mutation in CFF-16HBEge human bronchial epithelial cells and rescuing CFTR transcript and protein expression, suggesting this approach as a potential novel therapeutic option for CF patients. 125
Duchenne muscular dystrophy
Duchenne muscular dystrophy (DMD) is an X-linked genetic disorder resulting from mutations in the dystrophin gene, which lead to reduced or absent expression of dystrophin, a protein essential for muscle fiber integrity, and cause progressive muscle degeneration and weakness. 126 As a strategy for targeted correction of nonsense mutations in the dystrophin transcript, Katrekar et al. used two ADAR-based RNA editing systems: one employing a gRNA containing an ADAR2-recruiting region derived from GluR2 mRNA, coupled with a target-specific antisense region (GluR2–ADAR; Fig. 5A) and another consisting of a bacteriophage MS2 hairpin-bearing gRNA and the MS2 coat protein (MCP) fused to the ADAR2DD (MCP–ADAR; Fig. 5B). 83 Each of the systems was delivered via AAV8 into the mdx mouse model of DMD, which harbors a premature UAA codon (Q995X) in exon 23 of the dystrophin gene, resulting in RNA editing yields of up to 3.6% and dystrophin protein restoration of 1%–2.5%. 83 In addition, a CRISPR-based editing strategy was also employed to target a different nonsense mutation in the dystrophin gene. Specifically, Wang et al. introduced a compact and efficient RNA base editor (ceRBE), which replaces the relatively large dCas13 protein with an engineered 199-amino acid EcCas6e protein derived from class 1 CRISPR–Cas systems, fused to the ADAR2DD. 127 Its delivery via AAV9 into a humanized mouse model of DMD carrying the Q1392X mutation led to high A-to-I editing efficiency of approximately 68%, accompanied by a comparable ∼68% rescue of dystrophin expression. 127 Notably, ceRBE exhibited low transcriptome-wide off-target effects, which, combined with its compact size amenable to AAV packaging, highlight its value as a therapeutic RNA editor. 127
ADAR-mediated RNA editing tools for correcting the Q995X mutation in a Duchenne muscular dystrophy model.
Hurler syndrome
Hurler syndrome (HS), also known as mucopolysaccharidosis type I, is a rare genetic lysosomal storage disorder caused by a deficiency of the lysosomal enzyme α-
Ornithine transcarbamylase deficiency
Ornithine transcarbamylase (OTC) deficiency is a rare genetic disorder affecting the urea cycle, which is responsible for converting toxic ammonia to urea. 130 Mutations in the OTC gene result in complete or partial absence of OTC, leading to impaired ammonia detoxification and accumulation (hyperammonemia), which can cause neurological complications and potentially death if untreated. 130 Katrekar et al. reported the application of the exogenous GluR2-ADAR2 (E488Q) system, delivered via AAV8, in the male sparse fur ash (spfash) mouse model of OTC deficiency. 83 This model harbors a G>A point mutation at the last nucleotide of the fourth exon of the Otc gene, resulting in defective splicing. 83 The intervention yielded 4.6%–33.8% editing in the correctly spliced Otc mRNA and led to a partial (2.5%–5%) restoration of the OTC protein. 83
Parkinson’s disease
Parkinson’s disease (PD) is a progressive neurodegenerative disorder characterized by the loss of dopamine-producing neurons in the brain, leading to motor symptoms such as tremors, rigidity, and postural instability, as well as nonmotor symptoms, including cognitive decline, mood disorders, and autonomic dysfunction. 131 PD has been shown to be associated with mutations in the PTEN induced kinase 1 (PINK1) gene, which plays a crucial role in mitochondrial quality control, particularly through the regulation of mitophagy—the selective degradation of damaged mitochondria by autophagy.132,133 The Stafforst group utilized an editing vector containing ADAR2 and an R/G-gRNA against a recessive loss-of-function mutation in PINK1 (W437X), which is linked to some hereditary early-onset forms of PD. 76 The vector was cotransfected with a mutant human PINK1 W437X construct into PINK1-knockout HeLa cells, resulting in 10% RNA editing and restoration of PINK1/Parkin-mediated mitophagy 76 —a process impaired in PD—providing evidence for the effectiveness of this strategy in ADAR-mediated editing and functional rescue of PINK1.
Rett syndrome
Rett syndrome (RTT) is a rare neurodevelopmental disorder primarily affecting females and is most frequently associated with mutations in the MECP2 gene located on the X chromosome. 134 It leads to normal early development followed by regression in speech, motor skills, and hand use, along with other neurological impairments and comorbidities, such as anxiety, breathing irregularities, and seizures. 134 The Mandel group employed the λN-BoxB-ADAR2 system delivered by AAV to edit a severe human MECP2 G>A mutation (R106Q) in the DNA-binding domain.82,105 This approach achieved Mecp2 mRNA editing rates of approximately 72% in primary hippocampal neurons and 50% in mice, accompanied by restoration of MECP2’s ability to bind heterochromatin.82,105 They reported that, in vivo, targeted RNA editing was efficacious in nondividing cells and most prominent in the brainstem, with associated rescue of MECP2 expression and function, alleviation of Rett-like respiratory dysfunction, and extended survival, supporting the potential of the λN-BoxB-ADAR platform to ameliorate RTT symptoms in vivo. 106
Clinical Trials
With the rapid advancement of RNA editing technologies, ADAR-based RNA editing has progressed from preclinical proof-of-concept studies to early-stage clinical evaluation. Several therapeutic programs have now entered human trials, employing strategies that harness endogenous ADAR enzymes to achieve site-specific, RNA-level correction of disease-causing mutations, thereby setting the stage for transformative therapeutic modalities. Among the ADAR-recruiting platforms in clinical or near-clinical stages are WVE-006, WVE-008, KRRO-110, AX-0810, and AX-1412. 135
WVE-006 (Wave Life Sciences) is a GalNAc-conjugated oligonucleotide (AIMer) developed to treat AATD by recruiting endogenous ADAR enzymes to revert the pathogenic G>A mutation in SERPINA1 mRNA and restore the production of functional M-type (normal/wild-type) AAT protein. It is administered subcutaneously and leverages GalNAc conjugation to facilitate hepatic delivery.136,137 By reducing the accumulation of mutant Z-type AAT in the liver and restoring circulating M-type AAT, it aims to mitigate the liver and lung manifestations of AATD.136,137 Clinical development has progressed with the Phase 1 RestorAATion-1 study (ClinicalTrials.gov ID: NCT06186492) in healthy volunteers, which has been completed, 138 and the ongoing Phase 1b/2a RestorAATion-2 trial (ClinicalTrials.gov ID: NCT06405633) is evaluating WVE-006 in individuals homozygous for the Z allele (Pi*ZZ genotype) to assess safety, tolerability, pharmacokinetics, and pharmacodynamics. 139 Interim results from the 200 mg single and multidose cohorts and the 400 mg single-dose cohort of RestorAATion-2 indicate that WVE-006 restored M-type AAT to 64% of serum AAT while reducing Z-type AAT by 60%. 140 This AIMer was reported to be generally safe and well-tolerated, with all adverse events being mild to moderate in intensity and no serious adverse events observed. 140 Dosing continues in both the 400 mg multidose and 600 mg single-dose cohorts, with results from these cohorts expected in 2026. 140
WVE-008 (Wave Life Sciences) is an investigational GalNAc-conjugated AIMer targeting patatin-like domain 3, 1-acylglycerol-3-phosphate O-acyltransferase (PNPLA3) for the treatment of liver disease associated with the PNPLA3 I148M variant.140,141 This variant alters PNPLA3 function, leading to aberrant lipid metabolism and promoting fibrosis, hepatocyte ballooning, inflammation, and steatosis, yet no therapeutic interventions currently exist to directly target it.142–144 WVE-008 was designed to correct the I148M variant and restore normal protein function, thereby addressing the associated liver disease. Preclinical studies have demonstrated that RNA editing with the PNPLA3 GalNAc-AIMer restores functional PNPLA3 protein and decreases lipid accumulation.140,141 Clinical development planning for WVE-008 is underway, with a Clinical Trial Application (CTA) expected in 2026 for a first-in-human study enrolling homozygous I148M carriers previously genotyped, to assess safety, tolerability, pharmacokinetics, and pharmacodynamic endpoints.140,141
KRRO-110 (Korro Bio) is a synthetic RNA oligonucleotide developed using Korro Bio’s Oligonucleotide Promoted Editing of RNA (OPERA) platform, which, like Wave’s WVE-006, targets AATD by recruiting endogenous ADAR enzymes to correct the PiZ mutation in SERPINA1 at the RNA level.145,146 The REWRITE study (ClinicalTrials.gov ID: NCT06677307), a first-in-human Phase 1/2a clinical trial, includes both healthy volunteers and Pi*ZZ AATD patients and is designed to assess safety, tolerability, pharmacokinetics, and preliminary efficacy across ascending single- and multiple-dose cohorts.145,146 KRRO-110 is formulated in an LNP, which exhibits liver tropism, for intravenous infusion, with the potential to ameliorate both hepatic and pulmonary manifestations of AATD.145–147 Interim findings from the ongoing REWRITE study indicate that KRRO-110 has a tolerable safety profile. 148 All six planned single ascending dose healthy volunteer cohorts have been completed, and two AATD patient cohorts are under evaluation, with no serious adverse events or dose-limiting toxicities and mild-to-moderate infusion-related reactions observed, all of which resolved within 24 h. 148 However, because a single administration did not achieve projected levels of functional M-type AAT protein production, development has shifted toward a GalNAc-conjugated delivery approach, with a candidate nomination expected in 2026.
AX-0810 (ProQR Therapeutics) is a GalNAc-conjugated RNA editing oligonucleotide (EON) developed using the Axiomer™ platform by ProQR Therapeutics. 135 EONs harness endogenous ADAR enzymes to catalyze site-specific A-to-I editing within target RNAs. AX-0810 is designed to edit SLC10A1 mRNA, which encodes the sodium taurocholate cotransporting polypeptide, with the aim of modulating bile acid transport and reducing hepatocellular injury as a therapeutic strategy for cholestatic liver diseases. A CTA has been authorized, enabling initiation of a Phase 1 study in healthy volunteers, with dosing permitted to begin. 149
AX-1412 (ProQR Therapeutics) is another RNA EON developed using the Axiomer platform, designed to target beta-1,4-galactosyltransferase 1 (B4GALT1) mRNA, which encodes a key enzyme involved in glycoconjugate biosynthesis. 135 By introducing a protective variant into B4GALT1 RNA through site-specific A-to-I editing, it aims to reduce the residual risk of developing cardiovascular diseases. AX-1412 is currently in preclinical development with ongoing optimization for GalNAc delivery, and plans for clinical advancement are underway. 150
The field continues to expand, with additional companies actively pursuing therapeutic strategies based on ADAR-mediated RNA editing. For instance, AIRNA is developing a pipeline of RNA editing therapeutics utilizing its proprietary RESTORE+ platform, which enables A-to-I editing by recruiting endogenous ADAR enzymes. 151 The company’s lead candidate, AIR-001, is designed to correct the SERPINA1 PiZ mutation associated with AATD and is currently progressing toward clinical evaluation. 151
Conclusions and Future Perspectives
ADAR-based RNA editing systems offer a transformative, precise, and reversible approach to transcriptome modulation. By avoiding permanent changes to the genome, they present a safer alternative to DNA editing, with reduced risk of lasting off-target effects. Recent advances in gRNA design, delivery technologies, and strategies for recruiting endogenous ADAR have markedly improved in vivo performance. With early-stage clinical programs now underway, the field is entering a pivotal phase of therapeutic translation.
Building on this momentum, recent efforts have focused on improving the precision and efficiency of endogenous ADAR recruitment. Strategies such as chemically stabilized or circular gRNAs have enhanced editing durability and specificity, enabling effective editing in vivo, including in hard-to-reach tissues like the central nervous system. Parallel efforts in optimizing delivery vehicles, particularly LNPs and AAVs with refined tissue tropism, continue to broaden the therapeutic reach of these systems. However, the continued advancement of ADAR-based RNA editing will depend on addressing several key challenges and capitalizing on emerging opportunities that define the next stage of the field’s development.
One of the primary challenges in advancing ADAR-based RNA editing is enhancing editing efficiency and specificity in primary cells and in vivo contexts. Although editing in immortalized cell lines has generally achieved high efficiencies, replicating these outcomes in primary human cells and animal models remains a major hurdle. Innovations in gRNA engineering, including structure-guided design, circularization, and chemical modifications, are promising strategies. As an example of structure-guided design, Beal and colleagues leveraged structural insights from ADAR2–RNA complexes to develop the nucleoside analog 6-amino-5-nitro-3-(1′-β-d-2′-deoxyribofuranosyl)-2(1H)-pyridone (referred to as dZ) for placement at a guide strand position that contacts a conserved E residue (E488) in ADAR2. 152 This modification stabilizes an activated conformation of the enzyme–RNA complex, enhancing the catalytic rate of adenosine deamination by the wild-type ADAR2. 152 By mimicking the effect of the hyperactive E488Q mutant, the orphan base dZ-containing guide strand increased SDRE in human cells and mouse primary liver fibroblasts. 152 This study demonstrates how structure-guided modification of gRNAs can improve site-specific ADAR-mediated editing and advance the recruitment of endogenous ADARs for therapeutic RNA editing, with potential applicability to more clinically relevant systems, such as human primary cells and animal models. Additionally, efforts to localize endogenous ADAR isoforms or to design minimal exogenous ADAR variants with reduced immunogenicity could improve outcomes in tissues with low native ADAR expression.
Beyond improving editing efficiency and specificity, the field is actively exploring how to transcend canonical A-to-I editing. Efforts include the engineering or discovery of novel deaminases with alternative base preferences (eg, C-to-U, G-to-A) or the fusion of ADARs with other RNA-modifying enzymes to enable broader transcriptomic modulation. Likewise, base editors that can target splicing elements, UTRs, miRNA-binding sites, or RNA structural elements may enable nuanced control of gene expression, mRNA localization, and translation.
Complementing these approaches, recent advancements in synthetic biology have enabled the development of conditionally controllable RNA editors—tools that respond to signals such as small molecules, miRNAs, and specific cell states. These sophisticated platforms facilitate context-specific RNA editing, thereby broadening their potential applications to include cell type-specific therapies and dynamic transcriptome reprogramming. ADAR-mediated editing is also increasingly recognized as complementary to ASOs, RNA interference, and mRNA therapeutics. Combined approaches have the potential to enhance therapeutic durability and reduce dosage requirements by synergistically modulating multiple layers of post-transcriptional regulation.
Capitalizing on recent progress, the therapeutic scope of RNA editing is expanding beyond early efforts focused on well-characterized monogenic diseases such as AATD and HS. Future applications may extend to a broad spectrum of diseases, including various cancers driven by somatic mutations, neurodegenerative disorders, immune-related conditions, metabolic disorders, viral infections, and rare genetic syndromes. The integration of biomarker-guided patient stratification and advances in delivery technologies will be critical to broadening the clinical utility of RNA editors across diverse therapeutic areas.
In conclusion, ADAR-based RNA editing is no longer a conceptual frontier—it is rapidly becoming a therapeutic reality. As the field matures, a multidisciplinary convergence of RNA biology, delivery technology, and clinical translation will be essential. Continued innovation in enzyme engineering, gRNA design, and tissue-specific delivery will define the next wave of RNA therapeutics. With its unique advantages of reversibility, specificity, and programmability, ADAR-mediated RNA editing stands poised to reshape the landscape of nucleic acid therapeutics in the years ahead.
Footnotes
Author Disclosure Statement
The authors declare no conflicts of interest.
Funding Information
This research was supported by grants from the Korea Drug Development Fund (KDDF) funded by the Korea Government (MSIT, MOTIR, and MOHW) (RS-2025-02213647); the Korea Health Industry Development Institute (KHIDI) funded by the Korea government (MOHW) (RS-2025-25454991); the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2025-24535006, RS-2022-NR070845) the Korea Institute for Advancement of Technology (KIAT) funded by the Korea Government (MOTIR) (RS-2024-00435346); and Sookmyung Women’s University Research Grants (1-2403-2005).