Adenosine-to-inosine RNA editing in cancer: molecular mechanisms and downstream targets

Abstract

Adenosine-to-inosine (A-to-I), one of the most prevalent RNA modifications, has recently garnered significant attention. The A-to-I modification actively contributes to biological and pathological processes by affecting the structure and function of various RNA molecules, including double-stranded RNA, transfer RNA, microRNA, and viral RNA. Increasing evidence suggests that A-to-I plays a crucial role in the development of human disease, particularly in cancer, and aberrant A-to-I levels are closely associated with tumorigenesis and progression through regulation of the expression of multiple oncogenes and tumor suppressor genes. Currently, the underlying molecular mechanisms of A-to-I modification in cancer are not comprehensively understood. Here, we review the latest advances regarding the A-to-I editing pathways implicated in cancer, describing their biological functions and their connections to the disease.

Introduction

RNA, like DNA and proteins, can be modified by a variety of enzymes. RNA modifications impact RNA processing, stability, and translation, which have led to the recent development of a new field of research known as epitranscriptomics (the study of RNA modifications) (Roundtree and He, 2016; Saletore et al., 2012). To date, there are over 170 different types of RNA modifications that have been identified, and the development of high-throughput sequencing technologies has enabled the mapping of these modification in healthy and disease states (Boccaletto et al., 2018; Roundtree et al., 2017). All known RNA species, including messenger RNA (mRNA), ribosomal RNA (rRNA), transfer RNA (tRNA), microRNA (miRNA), and long noncoding RNA (lncRNA), have been shown to be modified. All four RNA bases and the ribose sugar could be targets of modification (Barbieri and Kouzarides, 2020; Li and Mason, 2014; Roundtree et al., 2017). Adenosine-to-inosine (A-to-I) RNA editing, in which adenosine (A) is converted to inosine (I), is one of the most prevalent RNA modifications. It was first detected over three decades ago in frog eggs and embryos as an activity that unwound RNA duplexes before researchers realized that adenosine residues in RNA have been converted to inosines (Rebagliati and Melton, 1987; Tan, 2023; Zinshteyn and Nishikura, 2009). This is followed by discovery of adenosine deaminase acting on RNA (ADAR) catalyzing A-to-I modifications (Wagner et al., 1989). Initially, a limited number of editing sites were discovered serendipitously in the protein-coding regions of mRNAs through comparing human genomic DNA versus cDNA sequences. With advances in next-generation and bioinformatic tools, it is possible to decipher A-to-I RNA editing globally, leading to a number of significant breakthroughs in the last decade. Surprisingly, the most frequent and widespread targets of A-to-I RNA editing are double-stranded RNAs (dsRNAs) made from inverted Alu repetitive elements (IR-Alu dsRNAs), which are located within introns and untranslated regions (Bazak et al., 2014; Peng et al., 2012; Porath et al., 2014). Moreover, inosine is frequently detected in tRNAs or viral RNAs, but seldom in DNA (Grosjean et al., 1996; Polson et al., 1996). Recent studies have demonstrated that A-to-I is mainly catalyzed by three groups of RNA adenosine deaminases. Apart from aberrant regulation of A-to-I regulators (ADARs) (Bass, 2002; Eisenberg and Levanon, 2018; Nishikura, 2010), adenosine deaminase tRNA-specific family (ADATs) catalyzes A-to-I conversion on tRNAs, whereas the testis-specific adenosine deaminase domain-containing (ADAD) protein family is essential for male fertility (Dai et al., 2023; Islam et al., 2023; Lu et al., 2023).

In this review, we will highlight the most recent developments in the field of A-to-I RNA modification, with a special emphasis on the roles of A-to-I in the genesis and progression of cancer, as well as the potential future directions of A-to-I research.

Regulators of A-to-I RNA modification

ADAR1/ADAR2/ADAR3

The family of ADARs was identified unintentionally by Bass and Weintraub (1988) as the enzymes responsible for “denaturing” dsRNA in Xenopus laevis embryos and unwittingly thwarting RNA interference (RNAi). Since then, ADARs (formerly referred to as DRADAs or DSRADs) have emerged as regulators of the gene expression output of a cell, and they are often deregulated in cancers.

In mammals, there are three ADAR genes, including ADAR1 (also known as DRADA), ADAR2 (also known as ADARB1), and ADAR3 (also known as ADARB2), with distinct differences in structure, location, and function (Shen et al., 2022; Xu and Öhman, 2018). ADAR1 has been demonstrated to be expressed and catalytically active in most tissues and encodes two isoforms generated from alternate promoter usage, a short, constitutive, and nuclear-restricted ADAR1 p110 isoform, and a longer, interferon (IFN)-inducible ADAR1 p150 isoform that is present in both the nucleus and cytoplasm (Hogg et al., 2011; Patterson and Samuel, 1995). ADAR2 has a more targeted expression pattern to tissues such as the brain, lungs, and arteries, and is responsible for the notably higher A-to-I editing rates in neuronal tissues (Baker and Slack, 2022). ADAR3, on the other hand, is specifically expressed in the central nervous system and has yet to show detectable editing action, although it may inhibit the activity of other ADARs in the brain (Chen et al., 2000; Oakes et al., 2017; Raghava Kurup et al., 2022; Tan et al., 2017; Walkley and Li, 2017; Zhang et al., 2018).

The functional domains of ADARs are shared. Each ADAR harbors two or three dsRNA-binding domains (dsRBDs) (~65 amino acids) with an α‑β‑β‑β‑α configuration that interacts directly with dsRNA (Stefl et al., 2006). The deaminase domain, which is crucial for its A-to-I editing activity, lies in the carboxy-terminal region. In addition, all ADARs contain a nuclear localization signal (NLS), which is responsible for the trafficking of ADAR1 to the nucleus and nucleolus; however, ADAR1 p150 isoform includes a nuclear export signal (NES) that allows it to shuttle between nucleus and cytoplasm (Desterro et al., 2003; Nishikura, 2016; Strehblow et al., 2002). Some structural characteristics are specific to individual ADAR members. For instance, ADAR1 has Z-DNA-binding domains (Zα and Zβ) (Herbert et al., 1997), whereas ADAR3 contains an Arg-rich single-stranded RNA (ssRNA)-binding R domain at its amino terminus (Chen et al., 2000). The deaminase domains of ADAR1 and ADAR2 can catalyze the hydrolytic deamination of adenosine-to-inosine in dsRNA. Inosine preferentially couples with cytidine (C) over uridine (U), and inosine is read as guanosine (G) by the ribosomes, thus enabling the same gene to gain the capacity to express various protein isoforms (Licht et al., 2019; Zinshteyn and Nishikura, 2009). Thus, ADARs increase protein diversity at posttranscriptional level, fine-tune regulation of gene expression, and enhance potential adaptability (Gommans et al., 2009).

Localization of ADARs is tightly regulated. Although ADAR1 p150 is generally localized in the cytoplasm and ADAR1 p110 is predominantly located in the nucleus, both of them can shuttle between the nucleus and cytoplasm (Desterro et al., 2003; Jarmoskaite and Li, 2024; Strehblow et al., 2002). Nuclear export of ADAR1 p150 is mediated by binding of the nuclear export factor exportin 1 (XPO1; also known as CRM1) to NES located within the Zα domain, together with Ran-GTP (Karki et al., 2021; Nishikura, 2016; Poulsen et al., 2001). Meanwhile, binding of transport protein 1 (TRN1) to the third dsRBD (containing NLS) mediates nuclear import of ADAR1 p110, a process inhibited by dsRNA binding; whereas nuclear export of ADAR1 p110 is regulated by XPO5–Ran-GTP and is mediated by dsRNA binding to dsRBDs (Fritz et al., 2009; Sakurai et al., 2017; Strehblow et al., 2002). As a consequence, dsRNA binding promotes a net efflux ADAR1 p110 from nucleus (Barraud et al., 2014). The predominantly nucleolar localization of ADAR2 is regulated by the binding of karyopherin subunit α1 (KPNA1) and KPNA3 to an Arg‑rich NLS in N‑terminal region (Ashley et al., 2024; Behm et al., 2017; Maas and Gommans, 2009). The nuclear localization and stability of ADAR2 are regulated by posttranslational modifications. Phosphorylation of Thr32 activates ADAR2 interaction with the prolyl‑isomerase PIN1 in a dsRNA‑binding-dependent manner, which isomerizes Pro33 and positively controls the nuclear localization and stability of ADAR2 (Eisenberg and Levanon, 2018; Keegan et al., 2023). In contrast, it has been observed that the E3 ubiquitin ligase WWP2 has a role in promoting the degradation of ADAR2 in the cytoplasm, which is why ADAR2 is usually not detected in the cytoplasm (Fig. 1A–C) (Marcucci et al., 2011; Vesely and Jantsch, 2021; Walkley and Li, 2017).

ADAT1/ADAT2/ADAT3

The A-to-I RNA editing processes have also been observed in tRNA. Methylinosine 37 (m¹I₃₇) is exclusive to eukaryotic tRNA^Ala and is generated by a two-step process, first involving the A-to-I modification catalyzed by homologous ADAT1 dimer, which belongs in the evolutionary clade containing ADARs at Position 37 in the anticodon loop, followed by methylation by tRNA methyltransferase 5 (Trm5) (Björk et al., 2001; Jühling et al., 2009; Maas et al., 1999). Methylinosine 37 modification is thought to repress translational frameshifts and enhance translational precision (Grosjean et al., 1996; McKenney et al., 2017).

Modification of A₃₄ of tRNAs is abundant in both bacteria and eukaryotes but nonexistent in archaea (Dixit et al., 2019). In bacteria, A₃₄-to-I₃₄ deamination reaction is present in two different tRNAs (mostly on tRNA^Arg_ACG and seldom on tRNA^Leu_AAG) and modification is catalyzed by the homodimeric tRNA adenosine deaminase A (TadA), the ancestor enzyme of all tRNA-specific deaminases (Rafels-Ybern et al., 2019; Wolf et al., 2002). In contrast, eukaryotic I₃₄ is found in eight tRNAs (tRNA^Leu_AAG, tRNA^Pro_AGG, tRNA^Ala_AGC, tRNA^Val_AAC, tRNA^Ser_AGA, tRNA^Ile_AAT, tRNA^Thr_AGT, and tRNA^Arg_ACG), and the A-to-I conversion is catalyzed by the heterodimeric adenosine deaminase consisting of catalytic subunit ADAT2 and tRNA-binding subunit ADAT3 (hetADAT). A conserved proton-shunting glutamate that is required for catalytic activity of ADAT2 has been lost in ADAT3, leaving this subunit inactive (Elias and Huang, 2005; McKenney et al., 2017).

Interestingly, inosine is the terminal modified base at Position 34, while at Positions 37 and 57 inosine could be further modified to a methylated state (m¹I₃₇, m¹I₅₇, or m¹Im₅₇) (Jühling et al., 2009; Machnicka et al., 2013). I₃₄-tRNAs are prone to internal cleavage by human endonuclease V, a highly conserved ribonuclease that specifically cleaves inosine-modified tRNAs at the anticodon (Vik et al., 2013). Stress conditions, such as oxidation and starvation, can trigger the cleavage of tRNAs within the anticodon loop as a part of the cellular response and the resultant fragments perform a variety of regulatory functions that are still largely unknown (Lyons et al., 2018; Phizicky and Hopper, 2010).

Adenosines at Position 57 (A₅₇) is converted into m¹I₅₇ in a two-step reaction. In the first phase, A₅₇ at TΨC-loop is methylated by an S-adenosyl-L-methionine (SAM)-dependent tRNA methyltransferase (TrmI) (Droogmans et al., 2003). Then, m¹A₅₇ is catalyzed to m¹I₅₇ by an enzyme that has not yet been identified (Grosjean et al., 1995). A₅₇ is also reported to be altered into di-methylated inosine (1,20-O-dimethylinosine, m¹Im₅₇) in hyperthermophilic species (Edmonds et al., 1991). Methylinosine 57 (m¹I₅₇ or m¹Im₅₇) is only observed in archaeal tRNA^Ile and its function is yet unknown (Fig. 1D–G).

ADAD1/ADAD2

The testis-specific ADAD protein family, including ADAD1 and ADAD2, are essential for male fertility. Specifically, ADAD1 is expressed in haploid spermatids, whereas ADAD2 is expressed in mid- to late-pachytene spermatocytes. Adad1^−/− male mice showed reduced sperm counts, decreased motility and malformed heads; deletion of Adad2 in mice resulted in male sterility, as Adad2^−/− germ cells are unable to progress beyond round spermatids (Connolly et al., 2005; Dai et al., 2023; Snyder et al., 2020). However, analysis of ADAD adenosine deaminase domain suggests that they are likely to be catalytically inactive. Thus, they might function as negative regulators of RNA editing in male germ cells.

Function of A-to-I editing in RNAs

ADAR-mediated editing has been identified at diverse RNA species, which could confer differentially effects on their function. A-to-I editing in the coding region can produce non-synonymous mutations, while A-to-I editing in the introns or 3ʹ-UTRs could modulate the expression levels of related coding regions. The majority of A-to-I editing sites are found in introns and 3ʹ-UTRs of coding genes, with 1% or less occurring in coding exons (Bahn et al., 2012; Picardi et al., 2015; Wang et al., 2013a). In ncRNAs, editing affects ncRNAs maturation and targeting, whilst A-to-I editing of tRNAs at wobble base at Position 34 is essential for decoding the redundancy of the genetic code.

Function of A-to-I editing in mRNA

A-to-I editing of mRNA can occur in exons, intron, and 3ʹ-UTR regions. There are millions of RNA editing sites in the human transcriptome. Only a very small fraction of those sites is in protein-coding mRNA sequences, with the vast majority in noncoding RNA sequences within untranslated regions and introns (Eisenberg and Levanon, 2018; Ramaswami and Li, 2014). Noncoding RNA editing sites in humans are most frequently observed in IR-Alu repetitive elements in which two adjacent Alu elements in opposite orientations in the same transcript form dsRNAs (Chen et al., 2008; Levanon et al., 2004; Ramaswami et al., 2012). Early mechanistic studies on ADARs mostly focused on protein recoding, which potentially alter amino acid sequence, thereby leading to inactivation or activation of the translated protein (Fig. 2A) (Datta et al., 2023). Where the coding sequence exhibits a double-stranded structure on the exon(s), these regions might be potential targets for ADAR1-mediated A-to-I editing. Several genes have been identified to undergo A-to-I editing in exon(s), such as antizyme inhibitor 1 (AZIN1), bladder cancer-associated protein (BLCAP), gamma-aminobutyric acid A receptor alpha3 (Gabra3), glioma-associated oncogene 1 (GLI1), and integrin a2 (ITGA2). Some intronic regions can also recruit and attach to ADAR1, leading to increased transcript stability and abundance in an A-to-I editing-dependent manner (Fig. 2B) (Amin et al., 2017; Liu et al., 2022). Moreover, A-to-I editing of intronic regions may result in various alternative splicing patterns, including creating donor sites (5ʹ GU splice sites), creating or destroying acceptor sites (3ʹ AG splice sites), or creating altogether new splice sites (Fig. 2C) (Lev-Maor et al., 2007; Mallela and Nishikura, 2012; Sakurai et al., 2010; Wang et al., 2020). A-to-I modifications in 3ʹ-UTR have two major effects on the regulation of gene expression. The first mechanism relies on regulating the stability of the mature mRNA. For example, ADAR1 recruits and interacts with human antigen R (HuR, gene name ELAVL1), a family of RNA-binding proteins (RBPs) selectively binds to single-stranded AU-rich RNA sequences to increase transcript stability (Fan and Steitz, 1998; Wang et al., 2013a). ADAR2 enhances mRNA stability by restricting the interaction with RNA-degrading proteins, such as HuR and PARN (poly(A)-specific ribonuclease) (Fig. 2D) (Anantharaman et al., 2017; López de Silanes et al., 2003; Meisner et al., 2004; Song et al., 2016). Another common one is editing of Alu dsRNA on 3ʹ-UTR, which otherwise binds to and is regulated by miRNA in the unedited state. Several studies have reported that RNA editing that occurs in the 3ʹ-UTR could create or destroy miRNA binding sites, thereby altering the mRNA stability of cancer-related genes (Fig. 2E and 2F) (Borchert et al., 2009; Liang and Landweber, 2007; Roberts et al., 2018; Soundararajan et al., 2015; Tomaselli et al., 2013; Yang et al., 2017; Zhang et al., 2016).

Function of A-to-I editing in noncoding RNA

A-to-I editing affects any of the miRNA biogenesis steps, including processing of primary miRNA (pri-miRNA) into precursor miRNA (pre-miRNA) by DORSHA, the conversion of pre-miRNA into mature miRNA by DICER, and loading of miRNA into RNA-induced silencing complex (RISC). Moreover, A-to-I editing can also impact miRNA target selection. In this way, the editing of certain miRNA precursors results in the altered expression or function of corresponding mature miRNAs (Fig. 2G) (Deiuliis, 2016; Slezak-Prochazka et al., 2010; Tomaselli et al., 2013). Additionally, lncRNA are targets of A-to-I modification. Editing primarily occurs in dsRNA regions on lncRNA by ADAR changes its structure, which affects binding of downstream target miRNAs. Long noncoding RNAs can also recruit ADARs, which edit mRNA and, consequently, might promote target mRNA degradation (Fig. 2H) (Flippot et al., 2019; Nishikura, 2016; Silvestris et al., 2020). Similarly, ADAR1 interacts with inverse complementary dsRNA regions (e.g., Alu repeats) of circular RNAs (circRNAs) to make its production less favorable (Chen and Yang, 2015; Ivanov et al., 2015). A-to-I editing can affect the binding of RBPs to flanking intron regions, resulting in changes in the production of circRNAs (Fig. 2I) (Shevchenko and Morris, 2018; Wang et al., 2023b).

Function of A-to-I modification in tRNA

tRNA translates the genetic code during protein synthesis and is crucial to the efficiency and accuracy of translation (Klinge and Woolford, 2019). tRNAs fold into a cloverleaf secondary structure and adopt an L-shaped architecture, where nucleobases at Positions 34, 35, and 36 form the anticodon that recognizes complementary codon triplets in mRNA (Kim et al., 1973). Eukaryotic ADAT deaminates A₃₄ in multiple tRNAs (eight tRNAs in humans). ADAT activity is important to eukaryotes, given the fact that most of the eukaryotic genomes lack genes that code for G₃₄-tRNAs. Functionally, the I₃₄-tRNA modification potentiates wobble-pairing flexibility of the anticodon, as I₃₄-tRNAs could recognize either A-, C-, and U-ended synonymous codons, whereas A₃₄-tRNA could only efficiently combine with U-ended codons (Crick, 1966). The major impact of A₃₄-tRNA A-to-I modification is on the efficiencies of protein translation, as codon composition bias and clustering of rare codons (codons with few copy numbers of cognate tRNAs) in regions of mRNAs limit the rate of translation (Kubo and Imanaka, 1989; Parmley and Huynen, 2009). Hence, the translation of genes rich in ADAT-sensitive codons [codons translated by I₃₄-tRNAs, amino acids threonine, alanine, proline, serine, leucine, isoleucine, valine, and arginine (TAPSLIVR)] can benefit from the increased decoding capacity of inosine-modified tRNAs (Chan et al., 2015; Lyu et al., 2020; Novoa and Ribas de Pouplana, 2012). In agreement with this prediction, self-renewing embryonic stem cells that express many genes enriched in ADAT-sensitive codons and also displayed enhanced ADAT2 expression (Bornelöv et al., 2019).

Role of A-to-I in cancer

Emerging studies have shown that A-to-I modification plays a significant role in cancer development. RNA adenosine deaminases, especially ADARs, are overexpressed in tumors and they aberrantly catalyze A-to-I modification, eventually influencing the expression of target genes involved in tumorigenesis or modulation of tumor microenvironment (Table 1). Below, we describe the significance of A-to-I editing in a variety of cancers (Figs. 3–4).

A-to-I editing of mRNA in tumorigenesis

ADAR1 is overexpressed in many cancers and primarily functions as an oncogene (Qin et al., 2014; Takeda et al., 2019) by promoting cell proliferation and migration (Dou et al., 2016; Sun et al., 2020). Multiple studies have demonstrated that hyper A-to-I editing of AZIN1, modulated by ADAR1, is associated with the tumorigenesis of colorectal cancer (CRC), endometrial cancer (Nakamura et al., 2022), gastric cancer (GC) (Wang et al., 2023a), liver cancer (Shibata et al., 2023), and esophageal squamous cell carcinoma (ESCC) (Qin et al., 2014), and that recoding of AZIN is correlated with worse prognosis. Mechanistically, A-to-I editing of AZIN1 transcripts led to a serine-to-glycine substitution at residue 367, which is localized to β-strand 15 and is predicted to cause a conformational change that promotes AZIN cytoplasmic-to-nuclear translocation and also increases its protein stability (Chen et al., 2013). AZIN1 functions as an oncogenic factor that enhances cell proliferation, invasion, and migration capabilities; and cancer stemness characteristics (Chen et al., 2013; Shigeyasu et al., 2018), thereby conferring gain-of-function phenotypes with augmented tumor-initiating potential and aggressive phenotypes. Beyond cancer cells, increased expression of AZIN1 through A-to-I editing enhances invasive potential of cancer-associated fibroblasts (CAFs) within the tumor microenvironment in colon, and is an important predictor of tumor invasiveness in CRC (Takeda et al., 2019). Alterative targets of ADAR1-driven recoding with gain-of-function have been reported. ADAR1 enhances the function of GLI1 through A-to-I editing that leads to a change from arginine to glycine at Position 701, which boosted GLI1 transcriptional activity and therefore pro-tumorigenic phenotypes in pancreatic ductal adenocarcinoma (Shen et al., 2021). In thyroid cancer (TC), ADAR1-mediated modification of CDK13 leads to a change from glutamine to arginine at Position 103. Edited CDK13 protein was found to be enriched in the nucleolus and it conferred enhanced cell proliferation and migration properties to TC cells (Ramírez-Moya et al., 2021).

Nevertheless, more frequently ADAR1-mediated A-to-I recoding events inactivate tumor suppressors to induce tumor progression. ADAR1 is a critical oncogene for triple-negative breast cancer (TNBC). Substitution of methionine by valine at residue 2,269 in filamin B (FLNB) due to recoding by ADAR1 abrogates tumor suppressive activities of the protein, leading to cell-cycle progression and invasion (Baker et al., 2022). In cervical cancer, ADAR1 modified the BLCAP transcripts at two sites (tyrosine-to-cysteine and glutamine-to-arginine) in its coding region-tyrosine-X-X-glutamine (YXXQ) motif that binds to Src-homology 2 domain of signal transducer and activator of transcription 3 (STAT3) and inhibits its phosphorylation. A-to-I recoding of BLCAP prevents its interaction to STAT3, leading to the activation of STAT3 signaling and cancer progression (Chen et al., 2017a). ADAR1 also functions as an oncogenic factor in multiple myeloma (MM). Mechanistically, NEIL1 (required for base excision repair) is a target of ADAR1 A-to-I editing. The editing of NEIL1 on exon 6 leads to lysine-to-arginine substitution at Position 242, and recoded NEIL1 protein demonstrates defective oxidative damage repair and loss-of-function characteristics, which consequently promotes the acquisition of sporadic mutations in MM (Teoh et al., 2018). In contrast to the prevailing notion that ADAR1-mediated recoding is largely pro-tumorigenic in nature, one report demonstrated that ADAR1p110-mediated A-to-I recoding of GABRA3 (isoleucine-to-methionine at Position 342) repressed GABRA3-mediated AKT signaling, leading to reduced cell migration and invasion in BC (Gumireddy et al., 2016).

Besides recoding, ADAR1-mediated A-to-I editing at 3ʹ-UTR contributes to tumorigenic phenotypes. ADAR1 A-to-I modified 3ʹ-UTR of MDM2 and prevents its targeting by miR-155. Increased MDM2 in turn compromises transcriptional activation of p53 and promotes CML development (Jiang et al., 2019). In CRC, ADAR-mediated RNA editing at 3ʹ-UTR of PVR could upregulate its expression by increasing the RNA stability, leading to tumor- and immune-related gene functions and pathways in CRC (Qian et al., 2024). ADAR1-catalyzed A-to-I RNA editing of 3ʹ-UTR of Rho GTPase activating protein 26 (ARHGAP26) mRNA abolished its pairing to miR-30b-3p and miR-573, allowing increased translation and protein expression of this oncogene in BC (Wang et al., 2013b). ADAR1 has also been shown to modify 3ʹ-UTR region of methyltransferase-like 3 (METTL3), an N⁶-methyladenosine (m⁶A) writer, and abrogates its interaction with miR-532-5p, resulting in increased METTL3 protein expression in BC cells (Li et al., 2022c). This, in turn, promotes m⁶A modification and translation of ARHGAP5, a driver of tumor progression and metastasis. Moreover, METTL3-driven m⁶A modification reciprocally boosts ADAR1 mRNA and protein expression (Tassinari et al., 2021), implying positive feed-forward circuitry between A-to-I editing and m⁶A modification in promoting tumorigenesis. Recent studies identified an A-to-I editing independent mechanism of ADAR1 involving 3ʹ-UTR. Phosphorylated ADAR1 (ADAR1p110) in the cytosol binds to 3ʹ-UTR of several anti-apoptotic genes, which prevents their binding to Staufen1 and subsequent mRNA degradation, thereby promoting cancer cell survival (Boulay et al., 2014; Crawford Parks et al., 2017; Gong and Maquat, 2011; Sakurai et al., 2017; Xu et al., 2015).

The role of intronic A-to-I editing in tumorigenesis remains understudied. One study has reported that A-to-I editing of focal adhesion kinase (FAK) at specific intron sites on chr8,141,702,274 increases the stability of FAK mRNA and expression of FAK protein to promote mesenchymal traits, migration and invasion of lung adenocarcinoma (LUAD) (Amin et al., 2017). In CML, A-to-I editing within the introns of glycogen synthase kinase 3β (GSK-3β) by ADAR1 caused the mis-splicing of GSK-3β with reduced capacity to inactivate β-catenin, thereby facilitating LSCs self-renewal via the activation of β-catenin signaling (Abrahamsson et al., 2009; Jiang et al., 2013). ADAR1-driven A-to-I editing thus has wide ranging pro-tumorigenic effects in multiple cancer types.

In contrast to ADAR1, ADAR2-mediated A-to-I editing is frequently associated with tumor suppression. In liver cancer, ADAR1 overexpression and ADAR2 downregulation predict poor prognosis and increased risk of postoperative recurrence (Chan et al., 2014), an effect attributed to the differential selectivity in mRNA targeting by ADAR1/2. Indeed, several tumor suppressors have been identified as targets of ADAR2. For instance, ADAR2 targets intronic Alu elements of phosphatase cell division cycle 14B (CDC14B), an upstream regulator of S-phase kinase-associated protein 2 (Skp2)/p21/p27 pathway, in glioblastoma (GBM) (Galeano et al., 2013). ADAR2-mediated A-to-I editing on CDC14B pre-mRNA increases its expression, with a consequent reduction of Skp2 and the induction of p21/p27-mediated cell-cycle arrest (Galeano et al., 2013). In core binding factor acute myeloid leukemia (AML), which is defined by cytogenetic abnormalities either due to translocation (8;21) (q22; q22.1) or due to inversion (16) (p13.1; q22), ADAR2 is down-regulated due to the expression of RUNX1-ETO additional exon 9a fusion protein that exerts a dominant negative effect on ADAR2 transcription (Darwish et al., 2023; Guo et al., 2023). This led to downregulation of ADAR2 targets coatomer subunit α (COPA) (isoleucine-to-valine substitution at reside 164) and the component of oligomeric Golgi complex 3 (COG3) (isoleucine-to-valine substitution at residue 635), both of which inhibit clonogenic growth. In ESCC, ADAR2 also had tumor suppressive function. ADAR2 promotes asparagine-to-aspartate change at residue 72 of SLC22A3, a novel metastasis suppressor in ESCC (Fu et al., 2017). Downregulation of SLC22A3 facilitates cell invasion and filopodia formation by abrogating the binding between SLC22A3 and α-actinin-4 (ACTN4). ADAR2 also modifies the insulin-like growth factor binding protein 7 (IGFBP7) mRNA and stabilizes its protein by altering protease recognition site of matriptase (lysine-to-arginine at residue 95), which is essential for IGFBP7-induced apoptosis and the inhibition of Akt signaling in ESCC (Chen et al., 2017b). In gastric cancer, ADAR2-induced recoded podocalyxin-like (PODXL) (histidine-to-arginine substitution at residue 241) protein was found to suppress tumorigenesis by neutralizing the tumorigenic capacity of wild-type PODXL. In this scenario, ADAR1 and ADAR2 were observed to exert mutual interference on the binding and A-to-I modification of PODXL, whereby overexpression of ADAR1 impairs the tumor suppressive function of ADAR2 (Chan et al., 2016; Thomas, 2016). On the contrary, other studies have described pro-tumorigenic function of ADAR2. In CRC, A-to-I editing of BLCAP by ADAR2 leads to a substitution from glutamine to proline at residue 5, accelerating degradation of BLCAP via the ubiquitination-proteasome pathway (Han et al., 2022). As BLCAP interacts with retinoblastoma 1 (Rb1) to prevent its inactivation, downregulation of BLCAP by ADAR2 accelerates G₁-S cell-cycle transition, increases cell growth, and inhibits apoptosis. Another study also demonstrated oncogenic function of ADAR2 in malignant pleural mesothelioma (PM), albeit such effects are independent of RNA editing function (Sakata et al., 2020).

Compared to ADAR1/2, few studies have shed light on the potential role(s) of ADAR3 in cancer. As it is devoid of A-to-I editing functionality, ADAR3 mediates its effect on cancer by inhibiting RNA editing, mostly notably that of ADAR2-mediated glutamate receptor ionotropic AMPA2 (GRIA2) A-to-I editing (glutamine-to-arginine substitution at residue 607) in glioma and glioblastoma (Oakes et al., 2017; Zhang et al., 2018). More work is required to uncover the exact role of ADAR3 in cancer. In summary, current evidence supports a largely oncogenic function of ADAR1-driven A-to-I editing in cancer, whereas ADAR2 might possesses tumor suppressive or pro-tumorigenic function in a cancer type-specific and context-dependent manner.

A-to-I editing of noncoding RNA in tumorigenesis

A-to-I editing primarily modulates targeting and maturation of noncoding RNAs, thereby affecting their roles in cancers. ADAR1 also promotes pre-miRNA cleavage by Dicer and loading of mature miRNA into RISC independently of its RNA editing function (Ota et al., 2013). A number of microRNAs have been shown to be edited by ADARs. A-to-I editing of miR-3144-3p (3_A < G) by ADAR1 in liver cancer drastically shifts the target specificity of this miRNA, with the defective target of its canonical mRNA target Musashi RBP 2 (MSI2, an oncogene), whereas generating gain-of-function binding to SLC38A4 mRNA (a tumor suppressor). This led to overexpression of MSI2 concomitant with the downregulation of SLC38A4, thereby promoting tumorigenesis (Kim et al., 2023). In CML progenitors, ADAR1-induced hyper-editing of compromises maturation of pri-miR-26a by preventing its cleavage by DROSHA, a tumor suppressor miRNA that inhibits cell-cycle progression. Down-regulated miR-26a therefore accelerated cell cycle and self-renewal (Jiang et al., 2019). In LUAD, ADAR1 mediates A-to-I editing of miR-381, a microRNA implicated in stemness, chemoresistance, and other cancer-relevant pathways (Anadón et al., 2016). In TC, the hyper-editing of miR-200b by ADAR1 attenuated the capacity of miR200b to bind to ZEB1, a major transcription factor driving epithelial–mesenchymal transition (EMT). Knockdown of ADAR1 thus suppressed ZEB1, EMT, and aggressivity of TC cells (Ramírez-Moya et al., 2020).

On the other hand, several studies have shown that ADAR1 editing of miRNAs contributes to an antimetastatic effect in melanoma. ADAR1 is frequently down-regulated in metastatic melanoma. ADAR1 silencing increased the expression of non-edited miR-455-5p, which promotes the metastasis through inhibition of tumor suppressor gene CPEB1 (Shoshan et al., 2015). In a similar vein, miR-378a-3p is edited by ADAR1 in nonmetastatic but not in metastatic melanoma cells, and the modified form of miR-378a-3p preferentially binds to 3ʹ-UTR of the alpha-parvin (PARVA) oncogene and suppresses its expression, thereby preventing progression of melanoma towards malignant phenotypes (Velazquez-Torres et al., 2018). Moreover, ADAR1 silencing impairs cell invasiveness in melanoma by downregulating β3-integrin at both posttranscriptional and transcriptional levels via paired box 6 (PAX6) and miR-22/miR-30a/d, respectively (Nemlich et al., 2018, 2020). Additionally, ADAR1 modulates miRNA processing in a RNA editing-independent manner by regulating Dicer, an enzyme that cleaves precursor miRNA, thereby promoting the biosynthesis and function of miRNA-149* and negatively correlated with the GSK3a expression in melanoma (Nemlich et al., 2013; Yujie Ding et al., 2020). These findings suggest that ADAR1 editing of miRNA might exert discordant effect on tumorigenesis in a cancer type- and stage-specific dependent manner.

Similar to that its role in editing mRNA, ADAR2-mediated miRNA editing has been linked with tumor suppressive function(s). In CRC, ADAR2 is directly phosphorylated by protein kinase C ζ (PKCζ), which promotes ADAR2 editing activity and is necessary to maintain miR-200, an EMT repressor. Depletion of ADAR2 led to decreased miR-200, EMT and liver metastases (Shelton et al., 2018). miR-376a*, one of the mature miRNAs from miR-376 cluster, has been shown to facilitate glioma growth, migration and invasion in unedited form. ADAR2-driven editing of miR-376a* inhibited oncogenic characteristics of this miRNA (Choudhury et al., 2012). Mechanistically, ADAR2-editing shifted the target specificities of miR-376a* from RAP2A (member of the RAS oncogene family) to autocrine motility factor receptor (AMFR), resulting in increased RAP2A together with downregulation of AMFR to mediate pro-tumorigenic and pro-metastatic effects. Global profiling of miRNA in GBM cells expressing wildtype or catalytic-dead ADAR2 revealed that ADAR2 loss-of-function induced the expression of ~90 miRNA, the majority of which are onco-miRNAs. ADAR2 editing of miR222/221 and miR-21 precursors, for example, reduces expression of mature onco-miRNAs in vitro and in vivo, resulting in impaired cell proliferation and metastasis of GBM (Tomaselli et al., 2015). Hence, ADAR2 editing of miRNA likely exerts tumor suppressive function in cancer.

Beyond its wide-ranging effects on miRNA expression and specificity, ADAR1 functions as a repressor of circRNA production, which is a consequence of A-to-I editing near reverse complementary matches, a structural element considered essential for circRNA synthesis (Ivanov et al., 2015). In hepatocellular carcinoma (HCC), Shi et al. (2017) demonstrated that androgen receptor (AR) transcriptionally activates ADARp110 isoform, which then downregulates circARSP91 (hsa_circ_0085154), a tumor suppressive circRNA. AR-activated ADAR1 RNA editing might thus contribute to sexual disparity in HCC. Several circRNAs, such as hsa_circ_0004872 (Ma et al., 2020) and circNEIL3 (Shen et al., 2021), have been shown to reciprocally regulate ADAR1 expression in cancer, suggesting complex interactive circuity between circRNAs and ADAR1-mediated editing in driving tumorigenesis.

Role of A-to-I in cancer therapy

A-to-I editing in chemotherapy and targeted therapy response

Chemotherapy is still the mainstays of the treatment of cancers; however, chemoresistance is frequent cause of poor prognosis. Studies have demonstrated that mRNA A-to-I editing adversely impacts chemotherapy in cancer. For instance, ADAR1 controls the expression of dihydrofolate reductase (DHFR), the molecular target of antifolate drug methotrexate, via A-to-I editing of miR-25-3p and miR-125a-3p binding sites in DHFR 3ʹ-UTR (Fig. 5A) (Nakano et al., 2017). In BRCA1-associated protein 1 (BAP1) wild-type mesothelioma, ADAR2 up-regulation enhanced the A-to-I editing of transcripts and 3ʹ-UTR. ADAR2 promotes DHFR expression and splicing of active folylpolyglutamate synthetase (FPGS), both of which drive resistance to antifolates. As a consequence, the depletion of ADAR2 sensitized mesothelioma cells to pemetrexed, a first-line antifolate chemotherapy for mesothelioma (Fig. 5B) (Hariharan et al., 2022). A recent study demonstrated that ADAR1-mediated A-to-I editing on the 3ʹ-UTR of stearoyl-CoA desaturase (SCD1) augments SCD1 mRNA stability in gastric cancer cells (Wong et al., 2023). Increased SCD1 promotes lipid droplet formation to alleviate ER stress and boost cancer stemness, leading to gastric cancer chemoresistance (Fig. 5C). Consistently, Yang et al. (2023) showed that the stabilization of ADAR1 protein by DEAD-box helicase 1 (DDX1) promotes malignant phenotypes and cisplatin resistance (Fig. 5D). These studies suggest ADAR1/2-mediated A-to-I editing as a potential therapeutic target in combination with chemotherapy.

ADAR1 is overexpressed in Anlotinib-resistant non-small cell lung cancer (NSCLC/AR). Knockdown of ADAR1 decreases expression of C-X3-C motif chemokine ligand 1 (CX3CL1) in NCI-H1975/AR and A549/AR cells after Anlotinib treatment. Exogenous CX3CL1 reverses the effect of ADAR1 deficiency on increased NSCLC/AR cell sensitivity to Anlotinib (Wu et al., 2022), implying that ADAR1–CX3CL1 axis promotes Anlotinib resistance (Fig. 5E). ADAR3, although devoid of A-to-I editing activity, could also confer resistance to alkylating agent temozolomide in GBM cells by activating nuclear factor-kappa B (NF-κB) (Raghava Kurup et al., 2022). Hence, ADAR1 overexpression in cancer could elicit resistance to drugs with differential molecular mechanisms of action (Fig. 5F).

A-to-I editing in immunotherapy response

Immune checkpoint blockade (ICB) therapy based on antibody-mediated blockade of key checkpoint molecules, such as cytotoxic T lymphocyte-associated protein 4 (CTLA-4) and programmed cell death protein 1 (PD-1), have achieved clinical success in some cancers. Nevertheless, many cancers are refractory to ICB therapy, prompting the need for potential adjuvants for improving its efficacy. One promising strategy to improve ICB response is to activate the innate immune system via pattern recognition receptors (PRRs) (Datta et al., 2023), leading to Types I and II IFNs in the tumor microenvironment that promote tumor control by directly inducing cell death or enhancing antitumor immunity. Double-stranded RNA, typically associated with viral infection, potently activates Type I IFNs via innate immune receptors, such as melanoma differentiation-associated protein 5 (MDA5), protein kinase R (PKR) and Z-DNA binding protein 1 (ZBP1) (Levanon et al., 2024). Besides, endogenous IR-Alu sequences or cis-natural antisense transcripts (cis-NATs) are sources of self-dsRNA that elicits type I IFNs response by activating PKR (Jiao et al., 2022; Kuriakose et al., 2016; Li et al., 2022b; Thapa et al., 2016; Upton et al., 2012). Z-RNA, a left-handed helix of dsRNA, binds to ZBP1 receptor-interacting protein kinase 3 (RIPK3) to induce caspase 8-dependent apoptosis, mixed lineage kinase domain-like protein (MLKL)-dependent necroptosis, and the nucleotide oligomerization domain-like receptor family pyrin domain containing 3 (NLRP3)-gasdermin D (GSDMD)-dependent pyroptosis (Fig. 2J) (Jiao et al., 2022; Kuriakose et al., 2016; Thapa et al., 2016; Upton et al., 2012). In this connection, ADAR1 isoform plays a gatekeeper role through A-to-I editing of endogenous IR-Alu elements in self-dsRNA and Z-RNA, which prevents self-dsRNA to activate PKR and ZBP1. Indeed, it has been proposed that IFN-inducible expression of ADAR1 acts as a negative-feedback mechanism to counteract the increased responsiveness to self-dsRNA during inflammation (Fig. 2K and 2L) (Chen and Hur, 2022; de Reuver and Maelfait, 2023). Concordantly, the deletion of ADAR1 in tumors cells led to elevated antiviral cytokines and chemokines in response to IFN stimulation. Sustained activation of dsRNA sensor pathway by ADAR1 deletion also reduces cancer cell viability, as shown by the application of anti-cancer epigenetic inhibitors capable of inducing the transcription of repetitive sequences that form dsRNAs (Fig. 5G) (Chen et al., 2021; Chung et al., 2018; Liu et al., 2019). Hence, a number of studies have explored whether ADAR1 blockade potentiates immunologic response in the context of ICB therapy.

Ishizuka et al. (2019) utilized an in vivo clustered regularly interspaced short palindromic repeats (CRISPR) screen for the unbiased identification of targets that synergize with ICB therapy and revealed ADAR1 as the top candidate gene selectively depleted in B16 murine melanoma in immunocompetent mice plus anti-PD-1 treatment as compared to mice lacking T cells. Mechanistically, ADAR1 depletion or mutation resulted in impaired A-to-I editing of dsRNA, leading to the activation of MDA5 and PKR receptors, which respectively promote immune infiltration and inhibit tumor growth. Loss of ADAR1 reversed anti-PD-1 resistance in antigen presentation-deficient, but IFN-sensitive, tumors by eliciting inflammation. Another study demonstrated that ADAR1 depletion triggered Z-RNA accumulation and activation of ZBP1–RIPK3-dependent necroptosis (Zhang et al., 2022b). As ADAR1 inhibitor is not available, Zhang et al. (2022b) identified a small molecule, curaxin CBL0137, which directly activates ZBP1 by provoking Z-DNA formation (Fig. 5H). In animal models of melanoma, CBL0137 reverses anti-PD-1 resistance by promoting ZBP1-RIPK3-dependent necroptosis and driving CD8⁺ T cell recruitment (Zhang et al., 2022b). ADAR1 also inhibits ZEB1-mediated PANoptosis (inflammatory cell death pathway) to promote CRC and melanoma in mice via its action on dsRNA (Karki et al., 2021).

Apart from attenuating dsRNA response, ADAR1 unleashes immunosuppressive lncRNA LINC00624 via A-to-I editing. LINC00624 suppresses major histocompatibility complex class (MHC) I antigen presentation and limits CD8⁺T cell infiltration in tumor microenvironment. Moreover, LINC00624 propagates a positive feedforward cycle by promoting stabilization of ADAR1 by inhibiting its ubiquitination-induced degradation via beta-transducin repeat-containing protein (β-TrCP), leading to resistance towards ICB and anti-human epidermal growth factor receptor 2 (HER2) treatment (Fig. 5I) (Zhang et al., 2022a). A-to-I editing-independent immunomodulatory effects of ADAR1 have also been reported. In GC, ADAR1 suppresses IFN signaling in GC through miR-302a-IRF9-STAT1 signaling thereby abrogating antitumor immunity (Fig. 5J) (Jiang et al., 2020). Lenalidomide is a highly effective drug against multiple myeloma by multiple mechanisms including immunomodulation, anti-angiogenesis (Kotla et al., 2009) and CRL4^CRBN E3 ligase inhibition (Fink and Ebert, 2015). ADAR1 was shown to induce immunomodulatory drug lenalidomide resistance in multiple myeloma by enhancing Alu-dependent editing and transcriptional activities of GLI1 (arginine-to-glycine substitution) (Lazzari et al., 2017), a Hedgehog (Hh) signaling pathway activator and self-renewal agonist (Fig. 5K). Taken together, ADAR1 is a promising immune therapeutic target for improving ICB therapy in cancer.

Targeting A-to-I modification

Pharmacological targeting of aberrant A-to-I modification via ADAR1 inhibition

Direct targeting of ADAR1

Multiple efforts have been made to develop small molecule inhibitors of A-to-I enzymes, especially ADAR1. A comprehensive high-throughput screening was performed to identify potential inhibitors of ADAR1 and two compounds, lithospermic acid and Regaloside B, which interact with ADAR1 Zα domain (Hong et al., 2024), the functional domain that binds to Z-RNA and Z-DNA. In another study, alendronate, etidronate, and zoledronate were shown to inhibit the ADAR1 Zα domain through interacting with Lys169, Lys170, Asn173, and Tyr177 of the Zα domain in a similar fashion to the helical backbone of Z-RNA (Choudhry, 2021). Short RNA duplexes incorporating nucleoside analog 8-azanebularine has also been shown to selectively inhibit ADAR1, but not ADAR2 (Fig. 5L) (Mendoza et al., 2023).

Targeting ADAR1 splicing and nuclear export

ADAR1 isoform switching into ADAR1p150, a highly active A-to-I isoform, is a potential target to inhibit ADAR1 activity in cancer. Rebecsinib is the first such inhibitor reported (Crews et al., 2023), and it was shown to bind the spliceosome core complex and impair ADAR1p150 activation. By inhibiting ADAR1p150-mediating A-to-I editing, Rebecsinib impaired the self-renewal of LSC in vitro, and prolonged survival of humanized mice harboring LSC in vivo. Moreover, Rebecsinib spared normal hematopoietic stem and progenitor cells, and it showed favorable toxicokinetic and pharmacodynamic properties.

An alternative strategy is to block the nuclear export of ADAR1p150, which predominantly resides in the cytoplasm to catalyze A-to-I conversion. As the N-terminus of ADAR1p150 contains a NES as mentioned above, treatments targeting nuclear export of ADAR1p150, such as selective nuclear export inhibitors (NEIs), have been reported to be beneficial in cancer treatment (Azizian and Li, 2020; Gravina et al., 2014). NEIs such as leptomycin B, Selinexor (KPT-330), and Eltanexor (KPT-8602) have antitumor efficacy in preclinical models. KPT-330 has recently received US FDA approval for relapsed/refractory multiple myeloma (Chari et al., 2019; Karki et al., 2021; Theodoropoulos et al., 2020). Additionally, the use of IFNs to induce ZBP1, a Z-RNA sensor normally sequestered by ADAR1, could exacerbate induction of cell death when ADAR1p150 is targeted by NEIs. Mechanistically, KPT-330 and KPT-8602 inhibition competes with ADAR1p150 and allowing induction of cell death via ZBP1. Consequently, targeting ADAR1p150/ZBP1 axis by the combination of NEIs and IFN therapy regressed tumors by inducing PANoptosis (Fig. 5M) (Serafimova et al., 2012; Vercruysse et al., 2017).

Harnessing ADAR1 for site-directed editing by CRISPR/Cas9

CRISPR-Cas systems have been widely employed to edit and modify specific nucleotides on DNA and RNA. Through taking advantage of the A-to-I regulators, a series of strategies have been proposed to manipulate A-to-I modification at specific RNA sites. Site-directed RNA editing by ADAR targets-specific transcripts by generating guanosine mismatches (Montiel-González et al., 2016; Yuan et al., 2023). Unlike DNA editing, RNA A-to-I editing does not impart a permanent modification to the host genome, which confers significant safety benefits and mitigates the side effects of off-target editing that may occur (Li et al., 2022a).

Although several site-directed editing strategies have been reported, they nonetheless share the same basic framework involving re-direction of ADAR deaminase domain to a specific region of the genome. To achieve site-directed RNA editing, a guide RNA (gRNA) is required to direct ADAR enzyme to the specific target site where editing is intended. The most common approach uses genetically engineered ADAR deaminase domain (ADARDD) that binds to a guide RNA. Exogenous ADAR proteins or their catalytic structural domains are fused to CRISPR-Cas13 (Tong et al., 2023; Xu et al., 2021), λ phage N protein (Austin et al., 2002; Baron-Benhamou et al., 2004; Keryer-Bibens et al., 2008; Montiel-Gonzalez et al., 2013), SNAPtag (Stafforst and Schneider, 2012; Vogel et al., 2014) or, and chimeric ADAR proteins (such as SNAP-ADAR) are then directed by sgRNA to the target sites to mediate A-to-I modification. These approaches differ in the mechanism that couple gRNA to ADARDD. For instance, CRISPR13b strategy utilizes catalytically dead Cas13b fused to N-terminal of ADARDD from ADAR1/2, which are co-delivered with gRNA composed of a 50 nt 5ʹ region complementary to the target sequence and containing a C mismatch to the targeted adenosine nucleotide. However, a pitfall in this approach is the need to deliver genetic engineered ADARDD-containing proteins, which might elicit undesirable immune responses in clinical applications. An alternative approach aiming to avoid the pitfalls of exogenous ADARDD by recruiting endogenous ADAR1/2 for targeted editing. Leveraging Endogenous ADARs for Programmable Editing of RNA (LEAPER) and CLUSTER are two systems developed to take advantage of endogenous ADAR, but their gRNAs differ. LEAPER is designed based on the premise that a long gRNA (~70 nt with a single C to A mismatch) could anneal target transcripts to form dsRNA substrates that are recognized by endogenous ADARs, whereas the gRNA for CLUSTER is a cluster guide RNA combining a specificity domain (20 nt with a single C to A mismatch), ADAR recruiting domain (R/G-motif), and a cluster of recruitment sequences (RS) (Katrekar et al., 2022; Qu et al., 2019; Reautschnig et al., 2022; Yi et al., 2022). Further modifications to sgRNA could enable light-triggered site-specific RNA A-to-I editing, leading to light-dependent point mutation of mRNA transcripts inside living cells and 3D tumorspheres in a spatially-defined manner. This represents a new approach for precise manipulation of A-to-I RNA editing in cancer treatment (Fig. 5N) (Zhang et al., 2023).

Future perspectives and conclusion

Many lines of evidence linking A-to-I editing aberrantly expressed to cancer strongly suggest that developing inhibitors targeting its pathways will be a fruitful pursuit. The A-to-I regulator ADAR1 is significantly overexpressed and promotes tumorigenesis in many cancer types, and high expression of ADAR1 often predicts poor survival in these patients, whereas ADAR2 is more frequently ascribed with a tumor suppressive role. Nevertheless, A-to-I RNA modification appears to serve as a double-edged sword in tumor development, the function of ADAR1 and ADAR2 is often cancer type-specific and context-dependent. More in-depth investigations in physiologically relevant models of cancer, such as ADAR1 or ADAR2 conditional knock-in/knockout mice, will unravel collectively impact of deletion of ADARs in vivo on tumorigenesis.

While our understanding of the biology of ADARs-mediated A-to-I editing on mRNA and noncoding RNA has rapidly expanded, the physiological and pathological significances of other A-to-I editing processes, such as the A-to-I editing of tRNAs, are only beginning to be explored. In humans, it has been reported that A-to-I modifications, both at Positions 34 and 37 of tRNA^Ala, are important for the recognition of autoantibodies generated against the anticodon stem loop of tRNA^Ala in patients suffering from myositis, a chronic inflammatory muscle disorder (Becker et al., 1999). However, their potential roles in tumorigenesis remain unexplored. The diversity of tRNA mutations and multiple diseases in which they are involved gives hope that this is only the beginning of the era of tRNA editing in cancer.

From a translational and clinical perspective, it will be of interest to further investigate the use of ADAR blockers for cancer treatment and combating chemoresistance, especially for ADAR1, in light of its predominant pro-tumorigenic effect in multiple studies. In particular, the combination ADAR1 inhibitors with immune checkpoint therapies to unleash antitumor immune response is another area of active investigation. Nevertheless, the development of ADAR-isoform-specific inhibitors is still underexplored. Additional preclinical studies will be needed to establish the efficacy of ADAR blockade for improving cancer therapy. Besides, whether ADAR expression or their downstream A-to-I modifications might be biomarkers for predicting cancer diagnosis or prognostication also deserve further investigations.

In conclusion, the A-to-I modification is involved in a variety of pathological processes, especially tumorigenesis. However, our understanding of A-to-I modification regulators is not yet comprehensive. Only two groups of deaminases complexes, ADARs and ADATs, have been identified to date, and there remain many questions regarding the intricate process of A-to-I modification. Firstly, it is unclear whether A-to-I editing is a dynamic and reversible process or whether there exist corresponding erasers that regulate the balance of A-to-I modification. Secondly, it is unknown whether ADAT2/3 regulates tumorigenesis by affecting the genome-wide codon use preference and protein translation efficiency of tRNA as mentioned above. Additionally, how many new viable codons can be generated by targeted mRNA editing? Do cells use this strategy to modulate protein diversity? There is no doubt that addressing these questions will require a great deal of work. However, the results generated from these experimental efforts will significantly advance our knowledge of the breadth of A-to-I function. Since the posttranscriptional network is intricate and various regulators are often connected, it is worthwhile exploring whether A-to-I and other posttranscription editing influence each other cooperatively to play a greater number of roles, especially in tumors. Further mechanistic studies are imperative to begin to unravel these mysteries.

Acknowledgements

All figures were created with bioRENDER.

Conflict of interest

The authors declared that there is no conflict of interests.

Funding

This project was supported by Health and Medical Research Fund, Hong Kong (22210122 and 08190706), RGC General Research Fund, Hong Kong (24100520, 14101321, 14101322, 14107924), Shenzhen-Hong Kong-Macao Science and Technology Program (Category C) Shenzhen (SGDX-20210823103535016), and RGC-Collaborative Research Fund, Hong Kong (C4008-23WF, C4039-19G).

Ethics approval and consent to participate

Not applicable.

Author contributions

H.C. organized and drafted the manuscript. J.Y. and C.C.W. revised the manuscript. All authors read and approved the final manuscript.

References