Background

RNA modifications play a crucial role in regulating various aspects of RNA function, including splicing, stability, localization, translation, and interactions with other RNA molecules and proteins. These modifications serve as key regulators of cellular physiology, and their alterations are closely associated with the onset and progression of numerous diseases, particularly malignant tumors [1, 2]. Several types of RNA modifications have been implicated in cancer development [3], such as N6-methyladenosine (m6A), N1-methyladenosine (m1A), 5-methylcytidine (m5C), N7-methylguanosine (m7G), pseudouridine (Ψ), and adenosine-to-inosine (A-to-I) editing [3,4,5]. Among these, A-to-I editing is especially prevalent in human cells. Research indicates that changes in A-to-I editing in cancer cells may promote tumor growth and confer resistance to apoptosis [6,7,8]. A-to-I editing can influence gene expression by altering mRNA splicing, RNA localization, microRNA (miRNA) binding, and transcript stability [9]. In addition, since inosine is read as guanosine (G) and preferentially pairs with cytosine (C) during translation, A-to-I editing can result in non-synonymous amino acid mutations, potentially altering protein function and contributing to cancer progression [10, 11].

A-to-I RNA editing

Adenosine deaminases acting on RNA (ADARs) enzymes are responsible for catalyzing A-to-I editing in double-stranded RNA (dsRNA). Three ADAR genes—ADAR1, ADAR2, and ADAR3—have been identified in the human genome, each containing two to three N-terminal dsRNA-binding domains and a C-terminal catalytic deaminase domain. ADAR1 exists in two primary isoforms: a 150 kDa isoform, which is induced by interferons and shuttles between the nucleus and cytoplasm, and a 110 kDa isoform, which is constitutively expressed and predominantly localized in the nucleus. The 150 kDa isoform binds to endogenous dsRNA, converting adenosine to inosine to prevent activation of melanoma differentiation-associated gene 5 (MDA5) and the downstream mitochondrial antiviral signaling (MAVS)-dependent type I interferon response. The 110 kDa isoform is primarily involved in nuclear RNA editing [12,13,14,15]. The ADAR2 gene is mainly expressed in the brain, where it is localized to the nucleus of neuronal cells and plays a key role in RNA editing in this tissue. All A-to-I RNA editing is catalyzed by ADAR1 and ADAR2. Although ADAR3 is structurally similar to ADAR2, alterations in key amino acids within its deamination domain and catalytic core prevent it from having catalytic activity. Instead, ADAR3 functions as a competitive binding agent for ADAR1 and ADAR2, thereby negatively regulating RNA editing [16]. Additionally, ADAR3 can bind to mRNA and regulate gene expression by influencing mRNA stability [17]. While most RNA editing sites are preferentially targeted by a specific ADAR, certain sites, especially those on miRNAs, lack preference for a particular ADAR [18]. A-to-I editing can also result from competitive or coordinated interactions between ADAR1 and ADAR2 [18,19,20]. In contrast, ADAR3 inhibits RNA editing and functions primarily as a monomer, while ADAR1 and ADAR2 can form both homodimers and heterodimers to regulate RNA editing processes [21].

ADAR acts on dsRNA to catalyze the site-specific deamination of adenosine to inosine (Fig. 1). During translation and RNA processing, inosine is recognized as guanosine (G) by transfer RNA (tRNA). In humans, A-to-I editing predominantly occurs in introns and untranslated regions (UTRs) of protein-coding genes [22], particularly in the 3’UTR, where editing sites are more densely distributed [23, 24]. ADAR enzymes primarily target repeat-rich sequences, such as Alu (SINE, short interspersed nuclear elements) elements and long interspersed nuclear elements (LINEs) [25, 26]. ADARs also act on dsRNA sequences longer than 20 base pairs (bp), with higher modification frequencies observed in sequences over 100 bp [27]. In fully base-paired dsRNAs exceeding 100 bp, only a few adenosines are selectively modified, whereas shorter, partially paired dsRNAs exhibit more selective modifications. This suggests that the secondary structure of the RNA plays a key role in determining the selectivity of A-to-I editing [28]. Some modification sites require adjacent long stem-loop structures, known as editing-inducing elements (EIEs), to recruit ADAR enzymes. Additionally, shorter and less stable double-stranded structures can facilitate efficient and selective editing by ADARs [29].

Fig. 1
figure 1

Diagram of RNA A-to-I editing. ADAR recognizes specific sites on dsRNA and deaminate Adenosine (A) to Inosine (I)

Full size image

A to-I editing levels are significantly altered in various malignant tissues, which can affect the occurrence and development of tumors (Table 1). Many tumors exhibit elevated modification levels [10, 30,31,32,33]. For example, breast cancer is characterized by increased ADAR1 expression and higher RNA editing in the 3’UTR of dihydrofolate reductase (DHFR) compared to adjacent normal tissues [34]. In endometrial cancer, increased ADAR1 expression and excessive RNA editing in antizyme inhibitor 1 (AZIN1) have been linked to poorer prognosis [35]. Similarly, elevated A-to-I editing of miR-411-5p contributes to the development of tyrosine kinase inhibitor (TKI)-resistant cells in patients with non-small cell lung cancer (NSCLC) [36]. In contrast, brain tumors often exhibit significantly reduced ADAR enzyme activity, reflected by decreased Alu RNA modification levels and lower ADAR expression. Both the increased A-to-I editing levels [35,36,37] and the dysregulation resulting from diminished editing activity contribute critically to tumorigenesis [38].

Table 1 Mechanisms of A-to-I editing in cancer progression
Full size table

Of note, in the same tumor, A-to-I editing can simultaneously exert regulatory effects through various mechanisms, such as inducing non-synonymous amino acid mutations, altering the immunogenicity of dsRNA, influencing miRNA maturation or targeting, and regulating RNA splicing. These diverse actions can result in either pro-tumorigenic or anti-tumorigenic outcomes, thereby shaping tumor progression. While the mechanisms and effects of ADAR activity vary across different cancers, increasing evidence suggests that ADAR1 generally promotes oncogenesis in many tumor types. However, in certain cases, such as melanoma and invasive breast cancer, RNA editing has been shown to exhibit specific antitumor properties [63, 64]. In breast cancer, A-to-I editing of GABAA receptor alpha3 (Gabra3) can inhibit the invasion and metastasis of breast cancer cells [63]. Additionally, RNA editing can increase the diversity of self-antigens presented by HLA, enabling recognition by the immune system and thereby enhancing antitumor immunity [64]. ADAR2 typically functions as a tumor suppressor gene [41, 54, 65]. For instance, in liver cancer, ADAR2 inactivates the PI3K/AKT/mTOR signaling pathway by producing a mutant protein that inhibits the expression of caveolin-1 (CAV1), contributing to the transformation of COPA from a cancer-promoting gene to a cancer suppressor gene [41]. In mouse breast cancer cells, ADAR2 inhibits circHif1a biosynthesis, allowing miR-195a-3p to interfere with P-glycoprotein (P-gp) translation, thus enhancing the susceptibility of cells to the chemotherapy drug adriamycin [65]. However, exceptions also exist. ADAR2-mediated RNA editing of SLC22A3 promotes tumor malignancy in esophageal cancer [48]. Meanwhile, in malignant pleural mesothelioma (MPM), knocking down ADAR2 suppresses MPM cell proliferation, motility, and invasion. Furthermore, the expression of ADAR2 lacking RNA-binding ability can inhibit tumor cell proliferation [66,67,68]. Meanwhile ADAR3, though less studied, appears to counteract ADAR1’s pro-tumorigenic effects, particularly in cancers like glioblastoma [69, 70]. Overall, the level of A-to-I editing is closely associated with patient survival and can significantly impact tumor drug sensitivity [33, 71, 72].

A-to-I editing regulates tumor progression by inducing non-synonymous amino acid mutation

ADAR edits the coding region of mRNA by deaminating adenosine to inosine, which is recognized as guanosine during translation. This results in changes in amino acid residue, altering the primary structure and function of the encoded proteins. One notable example is the editing of AZIN1 mRNA, which has been linked to various malignant tumors [10, 30, 35]. In hepatocellular carcinoma (HCC) and esophageal squamous cell carcinoma (ESCC), A-to-I editing changes serine to glycine at position 367 of the AZIN1 protein, leading to a conformational change that increases its affinity for antizyme. This interaction protects ornithine decarboxylase and cyclin D1 from degradation, thereby promoting tumor cell proliferation [10, 30]. In addition, elevated A-to-I editing of AZIN1 has also been observed in non-small cell lung cancer [32] and colorectal cancer [73]. The A-to-I-edited AZIN1 protein further drives tumor progression by upregulating interleukin-8 (IL-8), which in turn promotes angiogenesis, playing a key role in a variety of malignant tumors [35, 39, 73].

COPA (coatomer protein complex subunit alpha) is another key protein impacted by A-to-I editing [40, 41]. In metastatic colorectal cancer (CRC), A-to-I editing in the double-stranded RNA of COPA precursor mRNAs results in the substitution of isoleucine with valine at residue 164, producing the COPAI164V variant. This edited form of COPA promotes CRC metastasis by inducing endoplasmic reticulum stress [40]. In contrast, in hepatocellular carcinoma (HCC), ADAR2-catalyzed production of COPAI164V inhibits the PI3K/AKT/mTOR signaling pathway by repressing CAV1 expression, exerting a tumor-suppressive effect. When ADAR2 is down-regulated, insufficient COPA editing prevents the inhibition of CAV1 expression, leading to the activation of PI3K/AKT/mTOR signaling pathway, and ultimately promoting the development of HCC [41]. Another significant target of A-to-I editing is the bladder cancer-associated protein (BLCAP) gene, which encodes a protein that inhibits cell proliferation by promoting apoptosis in several cancers. In CRC, A-to-I editing of BLCAP increases the ubiquitination and degradation of the BLCAP protein and inhibits Rb1 phosphorylation. Due to A-to-I editing, the reduced BLCAP levels, diminish its tumor-suppressive function, by allowing tumor cells to transition from the G1 phase to the S phase more readily, thereby accelerating cell proliferation and reducing apoptosis [42].

In addition, A-to-I editing in the mRNA coding regions of various proteins can drive tumorigenesis and progression by altering their functions. For instance, A-to-I editing of the FLNB gene (filamin B) in breast and hepatocellular carcinomas reduces the tumor suppressor activity of the encoded protein, promoting tumor growth and invasion [10, 43, 44]. In thyroid cancer, A-to-I editing of CDK13 (cyclin-dependent kinase 13) increases its abundance in the nucleolus, promoting cancer cell proliferation and invasion [45]. Similarly, A-to-I editing in the RHOA (Ras homolog family member A) gene in lung adenocarcinoma enhance malignant cell proliferation and migration [11]. In intrahepatic cholangiocarcinoma, mutations in the KPC1 (cytoplasmic ubiquitin protein ligase 1) gene compromises the tumor-suppressive effects of the wild-type KPC1, thereby facilitating tumor growth and metastasis(Fig. 2) [46].

Fig. 2
figure 2

A-to-I editing in mRNA coding region resulting in missense mutation. When A-to-I editing occurs, as seen in AZIN1, it leads to non-synonymous mutations that promote tumor growth. This mutation elevates IL-8 levels, driving angiogenesis, while also increasing the protein’s affinity for anti-enzymes, leading to their degradation and supporting tumor proliferation, invasion, and metastasis. Similarly, A-to-I editing in the coding regions of other mRNAs, such as FLNB, CDK13, RHOA, and KPC1, causes inosine to be read as guanosine during translation. This substitution can result in changes to amino acid residues, potentially altering the structure and function of the proteins, thus facilitating tumor cell proliferation, invasion, and metastasis. Additionally, ADAR2-catalyzed A-to-I editing of COPA mRNA produces the COPAI164V variant, which promotes oncogenesis by inhibiting the PI3K/AKT/mTOR signaling pathway

Full size image

A-to-I editing regulates tumor progression by altering mRNA-miRNA interaction

MicroRNAs (miRNAs) are non-coding RNAs that primarily target the 3’ UTR of mRNAs to induce degradation or inhibit translation. A-to-I editing can influence miRNAs binding to mRNA by altering miRNA recognition sequences, splicing patterns, and miRNA maturation, thereby affecting tumor progression. A-to-I editing in either miRNAs or the 3’UTRs of mRNAs can disrupt their interactions [74]. Approximately 20% of A-to-I editing sites within the 3’UTR of mRNAs can alter miRNA targeting sites, impacting miRNA binding specificity. For instance, in osteosarcoma, the A-to-I editing site of 3’UTR is enriched in the miRNA-binding region of EMP2 and other oncogenes, thereby alleviating the inhibition of miRNA on target genes and increasing the expression of these oncogenes [49]. In breast cancer, ADAR1 modifies the mRNA of METTL3 (methyltransferase-like 3), altering its binding site for miR-532-5p, which increases METTL3 protein expression. This upregulation subsequently targets ARHGAP5 (Rho GTPase-activating protein 5), recognized by YTHDF1 (YTH domain family protein 1), promoting breast cancer cell proliferation, migration, and invasion [50]. Similarly, in glioblastoma, A-to-I editing in the 3’UTR of GM2A (GM2 ganglioside activator) is linked to the self-renewal of glioblastoma stem cells [51].

ADAR enzymes also directly edit miRNAs, impacting their role in cancer progression. In hepatocellular carcinoma, for example, ADAR1 modifies miR-3144-3p by substituting adenosine with guanosine at position 3 within the miRNA-binding domain, leading to the overexpression of Musashi RNA-binding protein 2 (MSI2), which promotes tumor growth [52]. In androgen-sensitive prostate cancer cells, ADAR modification of miR-379 weakens its growth-inhibitory effect, thereby facilitating tumor growth [53]. Conversely, ADAR2-modified miR-379-5p can induce apoptosis through CD97 (leukocyte differentiation antigen 97), inhibiting cancer cell proliferation across various tumor environments [54]. In thyroid tumors, excessive modification of the tumor suppressor miR-200b reduces its activity against its target gene ZEB1, leading to increased cancer cell invasiveness [55].

In the nucleus, ADAR enzymes compete with Drosha for binding to primary miRNAs (pri-miRNAs), thereby inhibiting Drosha’s ability to process pri-miRNAs into precursor miRNAs (pre-miRNAs) [75, 76](Fig. 3). Conversely, in oral squamous cell carcinoma (OSCC), ADAR1 forms dimers with Dicer, enhancing the modification and processing of oncogenic miRNAs. This activity plays a critical role in TGF-β-induced epithelial-mesenchymal transition (EMT), where elevated expression of ADAR1 promotes the maturation of oncogenic miRNAs, driving mesenchymal characteristics in OSCC cells [41, 77]. A-to-I editing also impacts viral miRNAs, such as those from Kaposi’s sarcoma-associated herpesvirus (KSHV), which is implicated in primary effusion lymphoma (PEL) [78].

Fig. 3
figure 3

Regulatory mechanisms of A-to-I editing in miRNAs and mRNAs. ADAR2 can promote CD97-mediated apoptosis by modifying miR-379-5p, leading to antioncogenic effects. Additionally, ADAR enzymes compete with Drosha for binding to pri-miRNAs, inhibiting their processing into pre-miRNAs. This inhibition enhances tumor cell proliferation, invasion, and metastasis. MicroRNAs typically function by binding to mRNAs to inhibit translation or promote degradation, thereby reducing the levels of target proteins. However, ADAR-mediated modifications of miRNAs (such as miR-3144-3p and miR-200b) or mRNAs (such as EMP2, MTTL3, and MSI2) can disrupt these interactions, allowing for the uninterrupted translation of oncogenic mRNAs. This process ultimately leads to increased tumor cell proliferation and invasion. Moreover, ADAR1 can dimerize with Dicer to enhance the modification and processing of oncogenic miRNAs, promoting epithelial-mesenchymal transition (EMT) and accelerating tumor progression

Full size image

A-to-I modified Alu regulates tumor progression by influencing coding sequence and Alu Immunogenicity

Alu elements are a type of SINE, formed by the fusion of heads and tails of 7SL RNA [79]. Most Alu elements are located within introns of protein-coding genes, as well as in the 3’ and 5’ UTRs, where they play a role in regulating gene expression [80]. When two Alu elements are positioned in opposite orientations and close proximity, they generate inverted Alu repeats (IR Alus), which promote the formation of dsRNA structures [81].

In malignant tumorigenesis, Alu elements within coding sequences are targets for A-to-I editing that can alter gene expression and lead to protein mutations. For instance, in lung adenocarcinoma, the RHOAiso2 splice variant, which contains an Alu-rich exon at the 3’ end of RHOA mRNA, undergoes A-to-I editing. This generates the RHOAiso2-R176G mutant protein, associated with dysfunctional RHOA activity, increased cancer cell proliferation and migration, and poor clinical outcomes [11]. Of note, IR Alus function as key post-transcriptional cis-regulatory elements and exhibit immunogenic properties [82, 83]. These elements can activate dsRNA sensors, such as TLR3, RIG-I, and MDA5, triggering IRF and NF-κB transcriptional pathways, and inducing the expression of interferons (IFNs), interferon-stimulated genes (ISGs), NF-κB-regulated genes, and pro-inflammatory cytokines [84,85,86]. In melanoma, AluJb, an innate immune activator, has demonstrated therapeutic potential by reducing tumor growth and extending survival when formulated as Alu-Np polymers [87]. Approximately 90% of A-to-I editing occurs on Alu RNA, with nearly all adenosine residues on Alu RNA having the potential to be edited [88]. ADAR1-mediated A-to-I editing can reduce the immunogenicity of Alu elements by disrupting their stem-loop structures, effectively converting aberrantly expressed “non-self” Alu sequences into “self” sequences. This mechanism may involve endogenous Alu transcripts, which would otherwise be recognized by MDA5 as non-self [49, 89, 90] (Fig. 4). In addition, A-to-I editing within dsRNA induces mismatches, leading to structural changes that hinder MDA5 recognition and autoinflammation. The inability to activate the inflammatory response allows tumor cells to evade the immune system, thereby avoiding cell death [91,92,93].

Fig. 4
figure 4

Regulatory mechanisms of A-to-I editing on Alu. The Alu-rich exon on the 3’ end of RHOA mRNA is a target for A-to-I editing, resulting in the production of the RHOAiso2-R176G mutant protein. This mutation enhances RHOA-GTP activity, leading to increased phosphorylation of the ROCK1/2 effector in lung adenocarcinoma (LUAD) cells, which promotes cell proliferation and migration in vitro. In vivo, this modification accelerates tumor growth in xenograft models and drives tumor progression in systemic metastasis models. Additionally, ADAR facilitates immune evasion by modifying Alu elements, thereby disrupting the immunogenicity of their stem-loop structures. This modification prevents recognition by dsRNA sensors such as MDA5, blocking activation of the interferon (IFN) signaling pathway and avoiding apoptosis

Full size image

CircRNA and lncRNA A-to-I editing levels and tumorigenesis

CircRNAs are a class of endogenous RNA molecules distinguished by their covalently closed loop structures, formed through back-splicing. This process involves the joining of a downstream 5’ splice donor to an upstream 3’ splice acceptor, often facilitated by RNA-binding proteins (RBPs) bound to flanking introns, as well as by base pairing of reverse complementary matches (RCMs) in these introns, such as IR Alus [94,95,96,97]. Unlike linear RNAs, circRNAs lack a 5’cap and a 3’poly(A) tail, which protects them from degradation by exonucleases and contributes to their increased stability [98]. Many circRNAs contain open reading frames (ORFs) and internal ribosome entry sites (IRESs), enabling them to encode functional proteins or peptides [99].

CircRNAs can function as “sponges” for miRNAs, indirectly regulating miRNA target genes by sequestering and competitively inhibiting miRNA activity, thus playing a regulatory role in various malignant tumors [100]. In pancreatic ductal adenocarcinoma (PDAC), circNEIL3 is produced through the interaction of ADAR1 with NEIL3. The circNEIL3/miR-432-5p/ADAR1/GLI1 (glioma-associated oncogene 1) axis promotes PDAC proliferation and metastasis by regulating the cell cycle and epithelial-mesenchymal transition (EMT). Additionally, circNEIL3 expression is negatively regulated by ADAR1 editing, ultimately influencing cell cycle progression and enhancing EMT in PDAC cells [59]. Notably, A-to-I editing can reduce the coding potential of circRNAs by altering codons and decreasing ORF/IRES scores [101]. In cervical cancer, ADAR1 enhances the binding of PTBP1(Polypyrimidine tract-binding protein 1) to the lateral intron of circCHEK2 by editing this intron, thereby promoting circCHEK2 biogenesis [60]. Additionally, A-to-I editing can impact the reverse splicing of circRNAs by altering the structure of Alu elements, thereby impacting the secondary structure formed by reverse complementary matches in circRNA introns [60, 101].

A-to-I editing is also associated with long noncoding RNAs (lncRNAs), which are noncoding transcripts longer than 200 nucleotides without significant protein-coding potential. LncRNAs are emerging as key regulators of various cellular activities, and changes in their expression or sequence have been linked to tumorigenesis and tumor progression [102]. For instance, the lncRNA MEG3 is highly edited in the normal cerebral cortex but shows reduced or absent expression in many cancers, including glioblastoma (GBM) [103]. Low MEG3 expression is correlated with shorter survival in GBM patients; reintroduction of MEG3 in glioma stem cells (GSCs) inhibits cell proliferation and tumor growth in vivo [104]. Conversely, antisense lncRNA PCA377 can specifically regulate A-to-I editing of coding genes by recruiting ADAR1, indicating that lncRNAs can selectively induce RNA modifications [61].

A study has demonstrated that A-to-I editing at specific sites on lncRNAs in humans, rhesus monkeys, mice, and flies can significantly impact the secondary structure of lncRNAs and their interactions with miRNAs, as revealed through systematic analysis [105]. In GBM cells, A-to-I editing of lncRNA is generally downregulated compared to normal brain cells [56, 62, 106]. Additionally, loss of A-to-I editing in lncRNAs can facilitate the release of cancer-promoting miRNAs (e.g., miR-331-3p) and miRNAs that bind tumor-suppressor mRNAs (e.g., miR-939-3p) [107, 108]. LncRNAs, in turn, can also modulate malignant tumor development by influencing A-to-I editing levels. For example, lncRNA FTX, which is highly expressed in colorectal cancer, promotes A to I RNA editing and miRNA synthesis by forming complexes with DICER and DHX9 proteins. This enhances DHX9-mediated A-to-I modification and promoting the growth of colon cancer cells [109]. Conversely, in leukemia, LNC-SNO49AB promotes ADAR1 dimerization at the post-translational level, increasing A-to-I editing throughout the cell and driving leukemia progression(Fig. 5) [34].

Fig. 5
figure 5

CircRNA, lncRNA, and A-to-I editing in malignant tumors. A. CircNEIL3 functions as a competitive inhibitor by binding to miR-432-5p, preventing its degradation and consequently increasing the expression of ADAR1. Elevated ADAR1 levels enhance A-to-I editing on GLI1, promoting epithelial-mesenchymal transition in pancreatic ductal adenocarcinoma (PDAC) cells. Additionally, overexpression of ADAR1 can inhibit the formation of circRNAs such as circNEIL3 and circSLC39A8, with the latter having an oncogenic role. ADAR-mediated A-to-I editing of circRNAs can affect their structure and coding potential, facilitating tumor cell proliferation, invasion, and metastasis. B. Downregulation of A-to-I editing on lncRNAs increases the binding affinity of lncRNA FTX to its target miRNAs, while reducing its interaction with miR-331-3p and miR-939-3p. This alteration promotes tumor cell proliferation, invasion, and metastasis. LNC-SNO49AB directly interacts with ADAR1, enhancing its homodimerization and boosting A-to-I editing activity. This results in increased intracellular A-to-I editing levels and contributes to leukemia progression

Full size image

A-to-I editing in malignant tumor treatment

In the treatment of malignant tumors, a patient’s biomedical characteristics can help predict potential outcomes and inform therapy. Studies have shown that the levels of ADAR and its substrates, such as miRNA editing levels [110], can not only help classify patients with malignant tumors but also indicate their resistance to certain drugs. This information serves as a valuable indicator for treatment decisions and prognosis, ultimately facilitating more targeted and effective therapies [111].

Some studies have explored A-to-I as a prognostic indicator in various malignancies, including gastric cancer and esophageal squamous cell carcinoma, suggesting that RNA editing features can help stratify patients based on their potential benefits from chemotherapy [48, 111, 112]. In lung squamous cell carcinoma, seven A-to-I editing sites were identified: TMEM120B chr12:122215052 A > I, HMOX2 chr16:4533713 A > I, CALCOCO2 chr17:46941503 A > I, LONP2 chr16:48388244 A > I, ZNF440 chr19:11945758 A > I, CLCC1 chr1:109474650 A > I, and CHMP3 chr2:86754288 A > I. Additionally, models predicting patient staging and survival based on these editing levels were developed [113]. In gastric cancer, a study identified 53 key editing loci with significant correlations by applying the Pearson correlation test between RNA editing levels and overall chemotherapy response [111]. This led to the development of a clinical staging and treatment prognosis test using the RNA editing (GCRE) signature as a malignancy indicator [111].

Studies have demonstrated that the level of ADAR1 in esophageal squamous cell carcinoma [48], the A to I editing level of miR-99a-5p in lung adenocarcinoma [114], and the level of ADAR2 in hepatocellular carcinoma are all strongly correlated with the clinical stage of these malignancies and patient survival [43]. Similar prognostic studies have been conducted across various cancers, including lung adenocarcinoma [115], glioblastoma [116], melanoma [117, 118], breast cancer [119], and prostate cancer [120], highlighting the potential of A-to-I RNA editing as a valuable biomarker for assessing disease progression and guiding treatment strategies [113].

Furthermore, ADAR plays a crucial role in immunotherapy. Studies have shown that ADAR1 deficiency increases the sensitivity of tumor cells to immunotherapy [121]. Inhibiting PKR or overexpressing either the wild-type or catalytically inactive mutant of the ADAR1-p150 isoform can partially rescue cell lethality induced by ADAR1 deficiency by blocking the PKR signaling pathway. This suggests that both catalytic and non-enzymatic functions of ADAR1 may contribute to preventing PKR-mediated cell death [122]. When ADAR1 is absent, A-to-I editing of RNA is reduced, allowing PKR and MDA5 to detect double-stranded RNA (dsRNA), thereby triggering tumor inflammation and inhibiting cell proliferation [121]. Consequently, ADAR1 deficiency can help overcome PD-1 checkpoint blockade resistance caused by the inactivation of antigen presentation in tumor cells [121].

Further research has shown that hnRNPC and ADAR deficiencies synergistically induce MDA5-dependent IFN responses [123]. Additionally, ADAR1 can inhibit the production of endogenous Z-dsRNA elements (Z-RNAs), which are enriched in the 3’ UTR of IFN-stimulated mRNAs. Knocking out or mutating ADAR1 leads to the accumulation of Z-RNA and activation of the Z-RNA sensor ZBP1, ultimately resulting in RIPK3-mediated necroptosis [124]. Specifically, in melanoma, silencing or overexpressing the ADAR1 isoform ADAR1-p110 decreases and increases T-cell migration, respectively. Antigen-specific T cells drive T-cell infiltration by releasing IFNγ, which in turn induces ADAR1-p150 expression [125].

Given the variations in ADAR levels across different tumors—both in expression and in their downstream editing sites—current research is increasingly focused on leveraging these differences as biomarkers for the detection and prognosis of malignancies. This approach highlights the potential of RNA modifications as a targeted tool in cancer diagnosis and treatment planning.

Conclusion and perspectives

Significant progress has been made in understanding how A-to-I editing regulates tumor progression. Numerous studies have detailed the mechanisms by which A-to-I editing affects tumor cells, including modifications in the coding regions of mRNAs that cause non-synonymous amino acid mutations, alterations in the non-coding regions of mRNAs that impact the immunogenicity of dsRNAs, changes to the structure of circRNAs, and effects on miRNA maturation and targeting through modifications in non-coding RNAs. These processes highlight the strong relationship between A-to-I editing and malignant tumors and their potential applications in clinical oncology. While A-to-I editing has diverse mechanisms of action across different types of tumors, there are also common targets, such as the editing of FLNB (Filamin B) and AZIN1 [10, 43] in hepatocellular carcinoma and esophageal squamous cell carcinoma, which influence cancer cell proliferation and metastasis. This reveals that while the effects of ADAR are tissue-specific, there are shared targets that could provide new avenues for the treatment and prognosis of various malignancies in the future.

Despite the multiple regulatory mechanisms of ADAR, its role in malignant tumors is not always clearly defined. For instance, while some studies suggest that A-to-I editing is unrelated to the development and progression of bladder cancer [126], others report contradictory findings. For example, recent research indicates that A-to-I editing levels in recurrent bladder cancer are associated with patient prognosis [127]. Thus, the precise relationship between A-to-I editing and tumor behavior remains ambiguous at times [128]. Additionally, the causal link between A-to-I editing and malignant tumor development is not yet clear. Although many studies have demonstrated that ADAR levels can affect tumor cell migration and invasion, other research, including mouse models of osteosarcoma, has shown that overexpression of ADAR1 alone is insufficient to initiate malignancy [129].

Despite the progress in understanding A-to-I editing, many specific issues remain unresolved. For example, ADAR1 mRNA is differentially expressed in various cancers, including renal refractory cells (KICH), renal clear cell carcinoma (KIRC), renal papillary cell carcinoma (KIRP), and hepatocellular carcinoma (LIHC). However, the underlying causes and mechanisms driving this differential expression remain unexplained [130]. In hepatocellular carcinoma (HCC), ADAR1 has been shown to be essential for cell survival under oxidative stress. It sensitizes HCC cells to oxidative stress by modulating the Keap1/Nrf2 pathway, leading to increased Keap1 and decreased Nrf2 expression. However, the exact mechanism of how it increases Keap1 expression has not been fully elucidated [131].

Questions also remain about the causes and mechanisms that lead to variations in A-to-I expression patterns across different malignancies [130]. Additionally, the mechanisms of A-to-I editing of circular RNAs and their roles in specific cancers are not fully defined. Further research in these areas could help identify new therapeutic targets for malignant tumors [132]. While early efforts to use A-to-I as a biomarker for malignant tumors have shown promise, the selection of relevant markers and predictive models requires further refinement before it can be fully implemented in clinical testing and treatment guidance.