ADAR1 and its implications in cancer development and treatment

Abstract

The family of Adenosine Deaminases Acting on RNA (ADARs) regulate global gene expression output by catalyzing Adenosine-to-Inosine (A-to-I) editing of double-stranded RNA and through interacting with RNA and other proteins. ADARs play important roles in development and disease, including an increasing connection to cancer progression. ADAR1 has demonstrated a largely pro-oncogenic role in a growing list of cancer types, and its function in cancer has been attributed to diverse mechanisms. Here we review existing literature on ADAR1 biology and function, its roles in human disease including cancer, and summarize known cancer-associated phenotypes and mechanisms. Lastly, we discuss implications and outstanding questions in the field, including strategies for targeting ADAR1 in cancer.

RNA modifications in human biology and disease

Post-transcriptional modifications to RNA have emerged as a widespread phenomenon and are described as the epitranscriptome of the cell. The field of epitranscriptomics (see Glossary), spurred on by advances in genomic sequencing and analysis, has vastly expanded and complicated our understanding of how the cell regulates its gene expression output. We now understand that over 100 naturally-occurring RNA modifications exist, providing an additional layer of information and regulation to a transcript’s sequence and structure. The majority of these are linked to structural RNAs, such as rRNAs and tRNAs, but several are found more broadly in mRNAs as well. In humans, the most predominant mRNA modifications studied include adenosine to inosine (A-to-I), cytidine to uridine (C-to-U), N⁶-methyladenosine (m6A), pseudouridine (Ψ), and 5-methylcytosine (5mC)[1].

The epitranscriptome describes a diverse group of RNA alterations. Base editing (A-to-I, C-to-U) involves irreversible deamination reactions to directly alter base pairing preferences, whereas methylation and other modifications (m6A, 5mC, Ψ) regulate RNA processing and stability via association with proteins and/or structural changes. RNA methylation (m6A, 5mC) is notable for involving writer, reader, and eraser proteins that allow for reversible and dynamic regulation[1], [2]. There is also interplay between different modifications, such as m6A inhibiting A-to-I editing in cis, which may contribute to the general lack of overlap between these marks[3].

Regardless of mechanism, these modifications have been demonstrated to play significant roles in human disease, as alterations to the enzymes responsible are linked to cancer, genetic birth defects, cardiovascular disease, neurological disorders, and more[1]. The m6A mark is known to regulate RNA metabolism including translation, microRNA (miRNA) processing, and splicing; consequently METTL proteins that catalyze m6A are implicated in cancers including AML, breast cancer, and glioblastoma via regulation of cancer gene stability[3]. The m5C methyltransferase NSUN2 is associated with defects in memory and learning, and its expression correlates with breast cancer and development[1].

RNA modifications represent yet another layer of regulation on global gene expression output, which can then have important implications for rewiring the cancer cell. Here, we focus on the inosine mark and one of its catalyzing enzymes, ADAR1, and examine their impact on cancer.

ADAR family members, isoforms & domains

The family of Adenosine Deaminases Acting on RNA (ADARs) was first discovered accidentally as the enzyme responsible for “denaturing” dsRNA in Xenopus laevis embyros and inadvertently thwarting RNAi experiments[4]. In more than two decades since its discovery, we have come to appreciate ADARs (previously called DRADA or DSRAD), and more broadly the field of RNA modifications, as a major regulator of the cell’s gene expression output. And, given the power of these modifications to alter the global transcriptome, dysregulation of these processes is increasingly implicated in human disease, especially in the process of malignant transformation to the cancer state.

ADARs are a group of enzymes that bind double-stranded RNA (dsRNA) as homodimers and catalyze the hydrolytic deamination of an adenosine nucleotide to form inosine. Adenosine-to-inosine (A-to-I) editing occurs only at regions of dsRNA and alters the base pair preference from guanosine to cytidine, thus functionally acting as an A-to-G mutation. There are three family members: ADAR1 (also called ADAR), ADAR2 (ADARB1), and ADAR3 (ADARB2). ADAR1 is ubiquitously expressed in nearly all tissues and contains two isoforms generated from alternate promoter usage. The p110 isoform is largely nuclear, while the p150 isoform is interferon-inducible and can be found in the nucleus and cytoplasm. ADAR2 has more targeted expression to tissues like the brain, lungs, & arteries, and is responsible for the especially high editing rates found in neuronal tissues. ADAR3 is brain-specific and has yet to demonstrate detectable editing activity, although it may suppress the activity of other ADARs in the brain[5].

Both ADAR1 and ADAR2 have similar domain structures that are critical for their enzymatic activity (Figure 1). Beginning from the C-terminus, the catalytic deaminase domain is essential for editing activity. Point mutations at a key glutamate residue, required for flipping the targeted adenine base out of a dsRNA duplex, are sufficient to inactivate the enzyme (E912A for ADAR1, E396A for ADAR2)[4]. Each ADAR contains two or three dsRNA binding domains (dsRBDs) that include a key KKxxK motif; mutation to EAxxA abolishes RNA binding activity, and consequently also editing activity. All ADARs contain a nuclear localization signal (NLS), but the ADAR1p150 isoform also contains a nuclear export signal (NES) that allows it to shuttle between the nucleus and cytoplasm. The p150 isoform contains an additional Z-DNA/RNA binding domain at its N-terminus, which may contribute to distinct binding preferences between the two isoforms. A recent analysis found that over half of ADAR1 editing sites were exclusively edited by ADAR1p150, possibly as a result of additional binding capabilities conferred by the Zα domain, although there was no significant difference in motifs surrounding p150-specific and p150/p110 sites[6].

Figure 1. Schematic of the domain structure of the family of human ADARs.

ADAR1 is composed of two isoforms, p150 and p110, which share identical sequence with the exception of additional N-terminal sequence for p150, including an additional Zα domain and the NES which allows for both cytoplasmic and nuclear localization. Both contain three dsRBDs that mediate dsRNA binding and homodimerization, both of which are critical for enzymatic activity. Mutation of the KKxxK dsRBD motif to EAxxA abolishes dsRNA binding and thus editing activity. The E912A point mutation in ADAR1p150 (E617A for p110, E396A for ADAR2) disrupts catalytic deaminase activity. ADAR3 is thought to be catalytically inactive, and contains a unique arginine-rich domain that confers ability to bind single-stranded RNA. Abbreviations: NES, nuclear export signal; dsRBD, dsRNA binding domain; NLS, nuclear localization signal; R, arginine-rich domain; Zα/β, Z-DNA/RNA binding domains.

A-to-I edits are predominantly found in noncoding regions and are most prevalent at Alu repeat elements interspersed throughout the human genome. When two similar Alu repeats are inverted in close proximity, transcription of this locus generates regions of strong complementarity that can form double-stranded RNA (dsRNA). Global analyses to identify A-to-I edit sites generally find that the large majority (80%+) overlap with Alu repeat elements, the largest source of endogenous dsRNA in the human cell[7]. Most A-to-I edit sites are generally found in introns and 3'UTRs, with 1% or fewer occurring in coding exons[8]-[10]. However rare, editing within coding regions does occur, and such examples were the focus of much early research into ADAR downstream mechanisms (e.g. GRIA2, AZIN1).

Analyses of ADAR editing preferences generally indicate nearest neighbor preferences for any base but G one nucleotide upstream (−1) and for G one nucleotide downstream (+1) (most commonly UAG), as well as a general depletion for A in the bases surrounding the edit[9]. Given the similarity in edit motif, there is significant overlap of editing targets for ADAR1 and ADAR2, with the subcellular location and expression level of ADARs dictating much of targeting preference, with the exception of p150-specific targeting noted above[10]. Algorithms for predicting editing based on sequence motif have shown limited utility (~75% accuracy for ADAR1) due to contributions from the deaminase domain, RBDs, and RNA structure in determining the likelihood of editing[11]. Thus, sequencing remains the most accurate and informative method for A-to-I editing detection.

ADAR mutant mouse models

The study of ADAR-knockout (KO) mouse models has contributed greatly to our understanding of the biology of ADARs and the relative contributions of different family members. Adar-KO is lethal at embryonic day 12.5, with mice displaying elevated IFN signatures and failed liver development[12]. Analysis of chimeric Adar-KO/WT mice demonstrated that Adar-null embryonic stem cells could contribute to development of murine brain, gonads, heart, kidney, and lung, but not bone marrow, liver, spleen, or thymus[12]. The embryonic lethality of Adar loss (or catalytic inactivation) is overcome by a second KO of Mavs or MDA5/Ifih1 (and delayed to E15.5 by Ifnar1 or Stat1 KO), indicating critical roles for dsRNA sensing pathway activation and the resulting innate immune response. However, all double KO mice still die shortly after birth, due to additional organ deficiencies not caused by MAVS pathway activation[13]-[15]. ADAR1p150-specific KO is also embryonic lethal but rescued by Mavs KO, implicating this isoform as the major mediator of MDA5/MAVS pathway lethality, whereas p110 contributed to organ abnormalities[14]. In agreement, later work found that ADAR1p110-specific KO mice survive to birth and displayed no activation of the MDA5/MAVS pathway[16]. However, these mice have high neonatal lethality rates that are completely rescued by crossing with catalytically inactive Adar, indicating that they are caused by an RNA-editing independent function of ADAR1p110. Adar KO in adult mice also led to a lethal MAVS-mediated inflammatory response and disruption of tissue homeostasis, confirming critical roles for Adar beyond embryonic development [14]. In short, Adar appears to play two key roles in development and homeostasis in mice: the editing-dependent action of ADAR1p150 in preventing activation of the MDA5/MAVS pathway, and the editing-independent role(s) of ADAR1p110 in maintaining organ development and homeostasis.

Adarb1-KO mice survive to birth but die within three weeks, and are rescued by knock-in of a single recoding Q/R edit site (with endogenous ~100% editing frequency) in the AMPA receptor Gria2[17]. Strikingly, a four-way mutant mouse (Adar^E861A/E861A, MDA5/Ifih1-null, Adarb1-null, Gria2^R/R) survives to adulthood and lacks any gross abnormalities, suggesting that aside from the Gria2 edit site and preventing MAVS activation, A-to-I editing is not essential for life[18]. However, a mouse model with targeted disruption of a recoding edit site in Flna demonstrates precursor signs of cardiac disease and defects in vascular cells, indicating that single edit sites may still influence biology in a subtler or more tissue-specific manner[19]. Finally, Adarb2-KO mice are grossly normal but display cognitive defects in learning and memory, and global A-to-I editing is essentially unchanged[20].

While KO mouse models allow for the study of whole-organism dynamics and remain a powerful tool, care should be taken in directly transferring conclusions from mouse to humans. Primates have over an order of magnitude higher editing rates than mice at conserved sites, largely attributed to the evolution of increased editing to accommodate primate-specific Alu repeat elements[21]. Especially given the relative ease of redirecting ADAR targeting by altering sequence or structure, it is conceivable that this conserved mechanism could be repurposed over evolutionary time to yield distinct functions in humans over mice. Additionally, the requirements for ADARs and their activities in normal homeostasis and development may not reflect dynamics in disease states – and thus a single edit site may still play an important role in cancer progression, even if it is not required for the maintenance of overall health and survival, as was suggested by the viability of the four-way mutant mouse model.

ADARs in human disease

In humans, mutations in ADAR1 have been linked to several type I interferonopathies, or diseases of dysregulated type I interferon homeostasis, in line with its observed role in suppressing IFN signaling. ADAR1 mutations are associated with Aicardi-Goutières syndrome (AGS), an autoimmune disorder characterized by aberrantly high interferon levels and various cranial abnormalities[22]. AGS-associated ADAR1 mutations cluster within the deaminase domain or the Z-DNA binding domains, which combined with the common pathways of nucleic acid processing and sensing regulated by other AGS-associated genes (IFIH1, RNASEH2A/B/C, etc.) implicates the A-to-I editing activity and masking of self-dsRNA of ADAR1 in the molecular pathophysiology of this disease[22]. Dyschromatosis symmetrica hereditaria (DSH) is an autosomal dominant inherited skin disorder characterized by hypo- and hyper-pigmentation of the face, hands and feet[23], [24]. Haploinsufficiency caused by mutations in ADAR1 is thought to cause DSH, and interestingly the mutations in DSH and AGS patients generally do not overlap.

Aside from discrete genetic perturbations in Mendelian diseases, disruptions in ADARs and A-to-I editing levels have also been associated with a wide variety of cancers, as well as several neurological and immunological diseases. ADAR2, being predominantly expressed in the brain, is commonly found to have reduced editing in brain disorders, including glioblastoma multiforme, Alzheimer’s disease, ALS, schizophrenia, and bipolar disorder[25]-[30]. ADAR1 mutations contribute to the cognitive defects of AGS by inducing a neuroinflammatory state and resulting encephalopathy[31]. Both ADAR1 and ADAR2 edit multiple recoding sites in the serotonin receptor 5HT_2CR to fine-tune its activity and downstream signaling in the brain[32].

Unsurprisingly given its connection to IFN signaling, ADAR1 has also demonstrated roles in regulating the virus-host interaction, either as a proviral or antiviral mechanism. Generally these roles involve A-to-I editing of viral genomes themselves or of host factors that regulate the antiviral response, although some editing-independent roles have also been described[33].

RNA editing dynamics have also been studied in the cardiovascular system, which also tends to have high A-to-I editing levels. ADAR1 editing of CTSS increases its expression via HuR-mediated stabilization, and increased CTSS expression is associated with atherosclerotic vascular diseases[34]. ADAR2 has also been implicated in cardiovascular disease through a recoding edit site in FLNA which, when disrupted, results in hypertension and cardiac remodeling in mice[19].

ADAR1 and cancer

Increased ADAR1 expression and/or activity has been implicated in many cancers, including hepatocellular carcinoma (HCC), non-small-cell lung cancer (NSCLC), gastric cancer, chronic myeloid leukemia (CML), esophageal squamous cell carcinoma (ESCC), colorectal cancer, oral squamous cell carcinoma, pancreatic cancer, multiple myeloma, cervical cancer, thyroid cancer, and breast cancer[7], [35]-[45]. Conversely, ADAR1 downregulation was associated with metastatic melanoma and kidney cancers[46], [47]. We have summarized prior literature that experimentally confirmed a role for ADAR1 in cancer hallmarks, including whether ADAR1 perturbation was tested in vivo and any downstream ADAR1 targets mediating cancer phenotypes (Table 1).

Table 1.

ADAR1 of cancer hallmarks

ADAR1 promoting cancer
Cancer	Hallmarks	In vivo	Genes involved	Ref.
Hepatocellular carcinoma	Growth, colony formation, invasion/migration	Y[35]	AZIN1[48], CircARSP91[72], ITGA2[73]	[35], [48], [72], [73]
Thyroid cancer	Growth, colony formation, invasion/migration	Y[45]	miR-200b-3p & ZEB1[45], CDK13[50]	[45], [50]
Cervical cancer	Growth	Y[74]		[63], [74]
Breast cancer	Growth, colony formation, migration/invasion	Y[65], [75]	FLNB, miR-27a-5p, miR-4485-3p[51], DHFR[56]	[56], [65], [75]-[78]
Chronic myeloid leukemia	Proliferation, self-renewal	Y[38], [52], [57], [79]	let-7[52], miR-26a and MDM2 & miR-155[57]	[38], [52], [57], [79]
Non-small cell lung cancer	Growth, metastasis, colony formation	Y[36], [80]	NEIL1 & miR-381-3p[36], FAK[81], AZIN1[82], CX3CL1[80]	[36], [80]-[82]
Gastric cancer	Growth & metastasis, colony formation, migration/invasion	Y[37], [83]	mTOR[83], miR-302a-3p & IRF9[84], circ0004872 & miR-224[85]	[37], [83]-[85]
Esophageal squamous cell carcinoma	Growth, colony formation, migration/invasion	Y[39], [86, p.2]	AZIN1[39]	[39], [86]
Colorectal cancer	Growth, stemness, migration/invasion*	Y[40]*	AZIN1[40], [87]	[40], [87]
Oral squamous cell carcinoma	Growth, colony formation, migration/invasion, stemness	Y[41]	DICER1[41]	[41]
Pancreatic cancer	Growth, colony formation, migration/invasion	Y[42]	c-Myc[42], circNEIL3 & miR-432-5p[88]	[42], [88]
Multiple myeloma	Growth, colony formation, self-renewal	Y[43], [49]	GLI1[43], NEIL1[49]	[43], [49], [89]
Prostate cancer	Growth	Y[90]	PCA3 & PRUNE2[90]	[90], [91]
Glioma	Growth, colony formation	Y[92]	CDK2[92]	[92], [93]
Melanoma	Growth	N	miR-149-3p & GSK3A	[94]
ADAR1 suppressing cancer
Cancer	Hallmarks	In vivo	Genes involved	Ref.
Melanoma	Growth & metastasis, invasion, T-cell killing	Y[46], [69], [95], [96]	miR-455-5p & CPEB1[95], miR-222 & ICAM1[97], miR-378a-3p & PARVA[98], ITGB3[96], [99]	[46], [69], [95]-[99]
Breast cancer	Migration/invasion*	Y*	GABRA3	[100]

^*indicates that phenotypes were only assessed by perturbing the target, not ADAR1 itself

ADAR mechanisms in cancer

ADARs can function to regulate the transcriptome in several distinct ways, and this is reflected in the diversity of ADAR1 targets and mechanisms that have been implicated in its pro-oncogenic effects. Most direct are effects mediated by A-to-I editing, which can alter RNA structure, binding motifs, coding sequence, and more to regulate the target. However, editing-independent roles have also been demonstrated for ADARs, as they can also function as an RNA-binding protein independently from catalytic activity. Such mechanisms are usually demonstrated by rescue of a given phenotype by the catalytic point mutant E912A (Figure 1).

Coding edits

Early mechanistic work on ADARs focused on the tantalizing, though rare, occurrences where editing resulted in direct recoding of a protein. One landmark study characterized a recoding edit site in antizyme inhibitor 1 (AZIN1) which resulted in an S367G substitution, thus altering the structure and localization of the protein to confer a greater pro-oncogenic function for the protein in HCC[48]. The K242R recoding edit in the DNA repair enzyme NEIL1 results in altered lesion specificity and less efficient DNA repair, which in multiple myeloma cells contributed to increased proliferation and colony formation[49]. The CDK13 Q103R edit site promotes increased growth, colony formation, and invasion in thyroid cancer and reduces the nuclear speckle localization of this splicing factor[50]. Filamin B editing yields an M2293V mutation that reduces nuclear localization and EMT-suppressive activity to promote invasion in triple-negative breast cancer[51].

microRNA mechanisms

ADARs can also bind and/or edit the hairpin intermediates of miRNAs. The primary miRNA (pri-miRNA) transcript is cleaved by nuclear Drosha/DGCR8 to yield the precursor (pre-miRNA), which is exported to the cytoplasm, cleaved by Dicer, and the mature miRNA loaded into the RNA-Induced Silencing Complex (RISC)[4]. ADARs have been demonstrated to regulate this process via A-to-I editing, dsRNA binding, or even RNA-independent direct protein interactions. An edit site in the seed region of miR-376a-5p yielded distinct binding preferences between the edited and unedited miRNA that resulted in opposing effects on invasion and migration in glioblastoma[25]. A-to-I editing of pri-let-7d was shown to inhibit processing to its mature form, thus promoting self-renewal capacity of CML progenitors to therapy-resistant leukemia stem cells[52]. Additionally, several studies have demonstrated a direct interaction between Dicer and ADAR1, independent of both catalytic and RNA binding activity of ADAR1, which can promote the processing of pro-oncogenic miRNAs[41], [53], [54].

3'UTR mechanisms

Given that the 3'UTR represents a large fraction of editing sites, it is unsurprising that ADAR1 targeting of the 3'UTR is another common mechanism for regulating gene expression in cancer. Editing of 3'UTRs can create or destroy miRNA binding sites to alter mRNA stability of important cancer genes, as several studies have demonstrated[55]-[57]. Binding of the 3'UTR of targets allowed ADAR1 to exclude Staufen1 binding and subsequent decay of anti-apoptotic genes, thus promoting stress-induced survival of cells in an editing-independent manner[58].

Splicing mechanisms

ADAR1-mediated splicing regulation can occur via edit sites that create/destroy splice signals, as well as binding competition with splice factors. Editing of HNRPLL in multiple cancer types was shown to promote exon 12A inclusion via SRSF1 recruitment, which may promote the expression of pro-growth genes like CCND1[59]. Editing-independent binding of ADARs was shown to regulate access of splicing machinery to multiple transcripts in ESCC, altering exon inclusion and function of several cancer-relevant targets[60]. Changes to global splicing patterns upon ADAR1 dysregulation has been documented in cancer cells, although only a limited number can be attributed to editing in cis, suggesting other mechanisms are responsible for the majority of alterations[61]. ADAR1 has also been shown to target trans splicing factors, such as CDK13 editing in thyroid cancer, which can then exert a larger influence on global splicing than individual edit sites[50].

R-loop edits

Several recent papers have demonstrated roles for A-to-I editing in the maintenance of genome integrity via editing of DNA:RNA hybrids. ADAR2-KD cells were found to have an impaired DNA damage response due to defects in clearing DNA:RNA hybrids[62]. ADAR1p110 was critical for proliferation of a subset of telomerase-activated cancer cells due to a requirement to edit telomeric variant R-loops to promote clearance and genomic stability[63]. Although these telomeric variants constitute a cancer-specific phenomenon, the authors also observed telomeric instability in Adar-null mouse embryonic fibroblasts. Unless a broader mechanism of R-loop clearance that is editing-independent exists, it is unclear how the editing-incompetent Adar^E861A/E861A, MDA5/Ifih1^−/−, Adarb1^−/−, Gria2^R/R mouse would have escaped such phenotypes. While the interplay between ADARs and the DNA damage response is not entirely understood, this suggests an additional pathway by which ADAR inhibition could constitute an effective therapeutic strategy.

As these examples illustrate, the multifunctional potential of ADARs allows for diverse possible mechanisms for transcriptome regulation, which can then contribute to cancer development. Thus it is critically important to assess the global transcriptome via unbiased sequencing within a given cancer context to understand how ADAR1 exerts its context-dependent effects on gene expression. Additionally, the potential impact of these mechanisms in a normal context is rarely assessed (the availability of primary cells being a common limitation) but represents an important consideration for understanding ADAR biology in full.

ADARs and immune signaling

As established in mouse models, ADAR1 and A-to-I editing play an essential role in suppressing IFNs and preventing an aberrant innate immune response, mediated by polymerizing of MDA5 on dsRNA and activation of MAVS. The interferon-inducible p150 isoform has been implicated as the major culprit in this pathway, presumably due to the editing of p150-specific targets it can access through its cytoplasmic localization and/or additional Z-DNA/RNA binding domain. Surprisingly, in ADAR1p110-null, ADAR2-null mice, only 2% of A-to-I edit sites in the brain are preserved by p150 activity and yet this is sufficient to prevent activation of MDA5[16]. Thus, it seems increasingly likely that MDA5 is sensitive to only a small subset of A-to-I editing targets, although the identity of these targets is yet unknown.

In addition to its role suppressing the MDA5/MAVS axis, suppression of PKR is another well-recognized facet to ADAR1 immune regulation. PKR is another antiviral dsRNA sensor that upon activation triggers translational shutdown via eIF2α phosphorylation and apoptosis. Loss of ADAR1 is generally associated with increased PKR phosphorylation, and baseline PKR expression has been hypothesized as a marker, or at least a requirement, for ADAR1 sensitivity [64]-[67].

Unsurprisingly, ADAR1 has also demonstrated roles in regulating antitumor immunity and synergizing with checkpoint blockade. In a screen for sensitizers to T-cell killing, ADAR1 emerged as a top hit in several murine cancers[68]. ADAR1-KO cells implanted in immune-competent mice had a strong synergistic response with anti-PD-1 treatment, overcoming several mechanisms of anti-PD-1 resistance and coinciding with increased levels of tumor-killing T and NK cells as well as immune cell-recruiting cytokines and IFNs[69]. This effect was dependent on Ifnar2 and Ifngr1 loss or STAT1 loss, indicating a requirement for IFN sensing, as well as for at least one functional dsRNA sensor (PKR or MDA5). Most compellingly, ADAR1-KO cells with antigen presenting defects (B2m-null were still sensitized to immunotherapy, suggesting the inflammatory tumor environment caused by ADAR1 loss is sufficient to restore antitumor immunity independent of CD8+ T-cell recognition.

However, the role of ADAR1 in cancer remains complex and context-dependent and consists of more than just its immune regulatory functions alone. Studies of ADAR1 perturbation in vitro or in immune-compromised mice have demonstrated roles in growth and metastasis, outside of roles in activating the antitumor immune response (Table 1). Also, rescue of ADAR1-loss-mediated growth defects in vitro by suppression of PKR or MDA5/MAVS are context-dependent and often partial, suggesting there are additional cell-intrinsic mechanisms regulating growth[65], [66].

Concluding Remarks and Future Perspectives

The diverse functions of ADAR1 in the cell complicate our understanding of its role in cancer. Between RNA binding, RNA editing, and protein interactions mediating its activity, ADAR1 exerts wide-ranging control over the global transcriptome. Yet, as the above examples have made clear, certain targets can cut through the noise to significantly impact cancer dynamics. In addition to regulating cancer mechanisms through specific gene targets, roles in suppressing immunostimulatory dsRNAs and, more recently, promoting genome integrity via R-loop clearance have emerged as additional angles to the story of ADAR1 in cancer.

While our understanding of ADAR biology has rapidly expanded, several key gaps remain that are relevant to the cancer context (see Outstanding questions). Unraveling these mysteries has major therapeutic potential to inform nuanced design of candidate drugs and offer better therapeutic precision. Depending on the mechanisms at play in a given cancer context, individual functional domains or isoforms could be targeted. Editing and/or p150-specific inhibition could be leveraged to induce an interferon response and cell death, and even synergize with other immunotherapies [69]. Precise suppression of a specific editing event could be achieved by altering the dsRNA structure using ASOs [70], [71]. Inducing telomere instability by p110-specific targeting was a proposed therapeutic strategy for telomerase-positive cancers, and we speculate that the broader role for p110 in the DNA damage response could lead to synergy with drugs in this pathway, such as PARP inhibition [62], [63]. Ultimately, a more complete ADAR1 targeting strategy may generate the most broad and robust response, but with increased risk for side effects in normal tissues. Finding the right balance for an optimal therapeutic window will be key to the development of anti-ADAR1 therapeutics.

Outstanding Questions Box:

• What is the identity of the RNA(s) responsible for activating MDA5/MAVS and PKR pathways? Emerging data suggests a key role for the ADAR1p150 isoform, possibly due to the unique binding properties of its additional Z-DNA/RNA binding domain. This would have major implications both for ADAR1-associated interferonopathies as well as targeting ADAR1 immune suppression in cancer.

• How could recently uncovered roles for ADAR1 in regulating genome integrity synergize with DNA damaging agents?

• If ADAR1 is ubiquitously expressed and its loss is generally deleterious in normal tissues as well, what is the potential therapeutic window for developing inhibitors to treat cancer and possibly other diseases?

Highlights Box:

The epitranscriptome represents an emerging novel layer of gene regulation that goes awry in cancer. Many RNA modifying enzymes have been implicated in cancer, including the Adenosine-to-Inosine (A-to-I) editors called ADARs.

ADARs are multifunctional proteins that regulate the transcriptome via A-to-I editing, RNA binding, and direct protein interactions. All of these functions have been implicated in cancer development in diverse tissue contexts. Here we focus on the predominantly oncogenic role of ADAR1.

In addition to mechanisms mediated by a specific target, ADAR1 also regulates interferon signaling via dsRNA sensors, which translates to an immunosuppressive function in the cancer context. Recently, a role in resolving R-loops has been proposed to regulate genome integrity and senescence, and thus another cancer hallmark.

Glossary Box:

Alu repeat elements

Short Interspersed Repeat Elements (SINEs) present in over 1 million copies in our genome due to retrotransposon activity. Alu elements in an inverse repeat orientation are responsible for the large majority of endogenous dsRNA, and thus are a major target of A-to-I editing.

Checkpoint blockade

A therapeutic strategy inhibiting one of several suppressive receptor-ligand interactions on T cells (checkpoints) in order to stimulate immune activity. Prominent examples include PD-1/PD-L1 and CTLA4/CD80 & CD86.

Epitranscriptomics

The collection of post-transcriptional RNA modifications that can alter the base pairing and/or activity of the modified RNA.

MDA5/MAVS pathway

An innate immunity pathway involving the sensing of long stretches of cytoplasmic dsRNA by MDA5 (part of the RIG-I-like receptor family). Binding and oligomerization of MDA5 on dsRNA allows for activation of mitochondrially-associated MAVS, which results in downstream type I interferon signaling via TRAFs.

microRNAs

A class of short ~22nt noncoding RNAs that direct Argonaute-mediated silencing of coding transcripts via sequence homology. The seed region (positions 2-7) is particularly critical and requires perfect complementarity with the target.

PKR

Protein kinase R, encoded by the eukaryotic translation initiation factor 2-alpha kinase 2 (EIF2AK2) gene, is another important component of the innate immunity pathway. PKR is activated by binding dsRNA and mediates translational suppression via eIF2α phosphorylation as well as cell death.

Z-DNA/RNA

Double-stranded DNA or RNA that is in the left-handed coil conformation.