ENDONUCLEASE V: FROM TRANSCRIPTOME REGULATOR TO CHEMICAL BIOLOGY TOOL

Abstract

Posttranscriptional modifications play a pivotal role in regulating the health and physiology of human cells and adenosine-to-inosine (A-to-I) editing is one of the most abundant of these modifications. Inosine is read as guanosine by the cellular machinery and this change impacts both the regulation and genetic code of RNA, in turn effectively rewiring both the transcriptome and proteome. Given these crucial functions, it is not surprising that A-to-I editing is implicated in a number of disease processes, including cancers and neurological disorders. While ADARs have been identified as the primary writers for inosine modification, putative readers and erasers have received far less attention. Recent studies have revealed that the Endonuclease V (EndoV) family of enzymes has the unique ability to detect and cleave inosine-containing RNAs, suggesting that EndoV may regulate the fate of edited RNAs in the cell. Despite this potentially important role, human Endonuclease V (hEndoV) remains relatively understudied, and in some cases, conflicting results have been presented regarding the activity of hEndoV. Here, we aim to bring together this existing knowledge in order to further spur investigation into this potentially crucial regulator of A-to-I editing. We first describe biochemical and structural studies that reveal the molecular aspects of inosine recognition and RNA cleavage, and the evolution of these functions from DNA repair in prokaryotes to RNA cleavage in eukaryotes. Based upon this knowledge, we discuss reports suggesting the role of hEndoV in human physiology. Finally, we highlight technologies that leverage EndoV for mapping and quantifying A-to-I editing in vitro and in cells and tissues.

Graphical Abstract

1. Introduction

RNAs undergo numerous post-transcriptional processing steps, including splicing, capping, polyadenylation, aminoacylation, and RNA nucleobase modification. More than 170 nucleobase modifications have been reported to date,^[1], with approximately 100 occurring in humans,^[2] each with its own potential set of writers which introduce the modification, readers which interpret the modification, and/or erasers which remove the modification.^[3,4] These modifications occur in all classes of RNA and specific modifications have been demonstrated to have myriad effects on RNA including, but not limited to, altering secondary structure, regulating RNA localization and stability, enhancing or suppressing translation, directing alternative splicing arrangements, and even recoding gene sequences.^[3,4]

Adenosine-to-inosine (A-to-I) editing is among the most common of these nucleobase modifications and is found in both DNA and RNA. In DNA, inosine is a form of damage which occurs via spontaneous deamination of adenosine as a result of nitrosative stress, such as exposure to superoxides and nitric oxide during inflammation.^[5,6] Additionally, the accumulation of deaminated noncanonical nucleoside triphosphates (such as dITP, dXTP and dUTP) in the cell’s nucleotide pool can lead to inosine misincorporation into DNA during replication.^[7] Due to its base-pairing properties, inosine is recognized as guanosine by polymerases and other cellular proteins, making it a promutagenic lesion that results in A-to-G transitions unless dealt with by DNA repair mechanisms.^[8]

While inosine in DNA is considered mutagenic, its presence in RNA is necessary and contributes to structural and functional diversity as well as transcriptome plasticity. Inosines are specifically introduced in RNA via highly regulated deaminases. For example, inosine is a key component of transfer RNA (tRNA). There are up to eight eukaryotic tRNAs that contain inosine at the wobble position, position 34, introduced by the enzyme adenosine deaminase acting on tRNA (ADAT).^[9] Alternatively, inosine is also commonly found in metazoan messenger RNA (mRNA), ribosomal RNA (rRNA), long non-coding RNA (lncRNA), and microRNA (miRNA) and is introduced by a family of adenosine deaminases acting on RNA (ADARs).^[10] In the coding region, A-to-I editing can result in codon change that alters the amino acid sequence of the translated protein. However, the vast majority of A-to-I editing events are found in non-coding regions of RNA, where they play important roles in regulating RNA secondary structure, splicing, localization, and stability.^[11] The alteration of secondary structure through A-to-I editing is one of the key functions of ADARs, as this prevents the accumulation of perfectly duplex double-stranded RNA (dsRNA) regions, which trigger an innate cellular immune response meant to combat RNA viruses.^[12] The incorporation of inosine in these long dsRNA stretches disrupts the perfect duplex, marking the RNA as “self” and preventing the activation of this immune response.

The Endonuclease V (EndoV) homologs are magnesium-dependent endonucleases that specifically recognize inosine bases and cleave the second phosphodiester bond 3′ to the deaminated base, generating 3′-OH and 5′-phosphate termini (Figure 1).^[13] The first EndoV homolog was isolated from Escherichia coli in 1977 and was initially characterized as an endonuclease with low hydrolytic activity on damaged DNA substrates.^[14] Thus, in prokaryotes such as E. coli, the main role of EndoV is to remove mutagenic inosines from the DNA. Because eEndoV only nicks the phosphodiester backbone on one strand and cannot excise the hypoxanthine base (as in base excision repair, BER) or remove the damaged base and surrounding nucleotides entirely (as in nucleotide excision repair, NER), it is thought to recruit additional enzymes to complete the repair process through an alternative excision repair (AER) pathway.^[15] Despite numerous studies on eEndoV, the full mechanism of inosine repair in E. coli remains poorly understood.

Figure 1:

Regulation of A-to-I editing in humans. ADAR serves as the writer for inosine modification and EndoV is capable of selectively cleaving edited RNAs.

The crystal structures of various EndoV homologs have revealed a conserved RNase H1-like fold containing a central beta-sheet comprised of eight beta strands and surrounded by seven alpha-helixes with a conserved catalytic center characterized by a highly conserved Asp-Glu-Asp motif.^[13] Due to their sequence and predicted structural homology, researchers initially thought that human Endonuclease V (hEndoV) would fulfill a similar role in human DNA repair and exhibit substrate specificity and cleavage activity on inosine-containing DNA akin to that of eEndoV. However, in 2013, two independent research groups demonstrated a major difference in substrate preference for hEndoV compared to eEndoV, specifically that hEndoV preferentially binds to and cleaves inosine-containing RNA rather than DNA. These reports further showed that hEndoV cleaves ssRNA more rapidly and to a greater extent than dsRNA, which we speculate is caused by the higher activation energy required for dsRNA cleavage, as the enzyme must disrupt the duplex structure to access the embedded inosine residue while the inosine and local backbone is more freely accessible in ssRNA regions. These studies revealed that hEndoV exhibits strong and specific endonucleolytic activity on RNA substrates containing riboinosine (rI), marking a functional divergence from its prokaryotic precursor.^[16,17] Consistent with its preference for inosine-containing RNA, the cellular distribution of hEndoV is mainly cytoplasmic.^[18] Even though the in vitro activity of recombinant hEndoV has been studied, its physiological function at the cellular level remains mostly unclear. Kong et al. developed the first mammalian EndoV knockout model and found that EndoV is not essential in mice, as EndoV^(−/−) animals exhibit normal development and lifespan. Moreover, global A-to-I editing levels were comparable between wild-type and knockout mice, suggesting that EndoV function is less likely to be directly coupled to global RNA editing turnover or degradation and may be subject to unique subcellular localization conditions that direct its cleavage activity.^[19]

This review aims to discuss the molecular evolution of hEndoV within the EndoV family, especially the divergence between prokaryotic and eukaryotic homologs, focusing on substrate recognition, catalytic mechanism, and metal ion dependence. Comparative structural analysis will also be presented to show how the catalytic motifs have adapted to support the enzyme’s substrate preference transition from DNA to RNA. Furthermore, we will explore the current understanding of the localization, isoform diversity, potential cellular functions of hEndoV, and discuss its potential biological implications in disease. Finally, we will describe recent technological advances and commercial applications that use the inosine-recognition capacity of EndoV to enable sensitive detection and spatial visualization of A-to-I editing events.

2. Characteristic Structural and Mechanistic Features

The Endonuclease V enzymes share the ancient-in-origin RNase H1-like fold with several other nucleases, including the RNA-DNA hybrid nucleases RNase H1 and Argonaute and DNases including HIV-1 integrase, UvrC, RAG1/2 recombinase.^[20] Substrate specificity of such nucleases is determined by the peculiarities inherent in each enzyme, and EndoV orthologs can vary with regard to their ability to cleave DNA or RNA despite sharing much or their sequence, being nearly superimposable structurally, and sharing a common mode of inosine recognition and cleavage.^[20] The unique structural features which confer the ability to specifically recognize inosine in nucleic acids include a wedge-shaped motif which stabilizes a single stranded-double stranded junction and facilitates base-flipping, inosine recognition in the binding pocket which holds the flipped base, and the site and mechanism of catalysis, all of which are discussed in detail in this section.

2.1. The Recognition Motif

Dalhus et al. solved the crystal structure of EndoV from Thermotoga maritima (TmEndoV), yielding the initial molecular insights into inosine recognition and backbone cleavage common to all EndoV orthologs.^[21] A wedge-like motif formed by the PYIPGL/F (79–85 in TmEndoV) residues in prokaryotes and PYVSGF (90–96 in mouse EndoV (MmEndoV) and hEndoV) residues in eukaryotes plays a crucial role in deforming the duplex and flipping the hypoxanthine base 90° at the wedge-motif towards the minor groove and into the binding pocket. A highly conserved tyrosine residue (Y80 in TmEndoV, Y91 in MmEndoV and hEndoV) then occupies the flipped base position, simultaneously base stacking with the nucleobase 3’ to the inosine and hydrogen bonding to the backbone between the bases (Figure 2B).^[22] The flipped hypoxanthine occupies the EndoV binding pocket and is stabilized through hydrogen bonding with protein backbone amide bonds (Figure 2C). Because the inosine binding pocket in the substrate-bound form is largely stabilized by backbone hydrogen bonds to the base, there is little impact on the shape and makeup of this pocket across species despite some sequence divergence between prokaryotic and eukaryotic homologs. Molecular dynamics simulations suggest a peculiar effect that contradicts the role of tyrosine in the wedge motif. The mouse EndoV can transition between two different conformations, with a wider binding cleft and Y91 side chain rotamers pointing both into and away from the RNA groove, conferring plasticity to the wedge motif not captured in crystal structure.^[23] These slight structural differences may also be contributing to differences in substrate specificity—whereas inosine substitutions found in DNA are typically random and isolated, inosines introduced into RNA can be clustered in hyperedited regions, potentially requiring different binding modes for efficient recognition of preferred targets in mammalian EndoVs compared to prokaryotic homologs.

Figure 2:

A) Crystal structure of mouse endonuclease V (MmEndoV) in complex with DNA substrate (PDB: 6OZL), B) The wedge-motif representing conserved PYVSGF amino acid residues, Tyr91 occupying the hole created by the flipped base, C) Stabilization of the flipped hypoxanthine base through backbone protein interactions D) The conserved catalytic residues and metal ions at the cleavage site, Glu100 interacting with ribose 2’-OH group

2.2. Catalytic Residues

Endonuclease V enzymes have conserved DDED/DDEH tetrad positioned 3’ to the hypoxanthine base binding site for the scissile phosphodiester bond hydrolysis in the active site.^[22,24,25] These amino acid residues are involved in metal coordination to facilitate the cleavage reaction through the coordination and polarization of water molecules. In addition, the glutamate residue of the tetrad is found to interact with the hydroxyl group of the ribose sugar to impart RNA cleavage preference for these enzymes over DNA.^[22] Active site mutants are severely compromised in their cleavage activity. For example, the mutants, D43A, E89A, D110A in TmEndoV^[24] and D52A, Y91A, and E100A in human EndoV show negligible cleavage activity, though many of these mutants retain their ability to bind inosine containing nucleic acids.^[25]

2.3. Metal Ion Dependence

EndoV phosphodiester cleavage is divalent metal mediated, natively using two Mg²⁺ ions bound by the catalytic tetrad to coordinate and polarize water molecules for hydrolytic cleavage.^[26] Magnesium dependence has been demonstrated through in vitro cleavage assays on nucleic acid substrates with recombinant EndoV, and both prokaryotic and eukaryotic EndoVs have been shown to also exhibit cleavage–and to do so with increased activity–in the presence of manganese, potentially due to its unique polarizability.^[27,28] However, the presence of calcium inhibits the cleavage reaction while preserving binding ability due to the increased size of the calcium ion imparting imperfect geometry to accommodate the cleavage reaction.^[29] Even though metal dependance is a common characteristic, the number of metals involved in catalysis varies depending upon the origin of EndoV.^[30] The presence of a single metal ion in the crystal structure of TmEndoV suggests that the prokaryotic EndoV can perform a single metal dependent catalysis.^[21] On the contrary, the crystal structures of eukaryotic EndoV contain two metal ions.^[23,26] Furthermore, a third transient metal ion was found to be necessary through in crystallo cleavage reactions carried out using MmEndoV, which is similar to the mechanism observed for both RNase H1 and DNA Pol η via similar methods.^[22]

2.4. The Cleavage Mechanism

Prokaryotic and eukaryotic EndoV catalytic mechanisms differ due to the differential metal dependance. Wetmore et al. proposed crucial steps in phosphodiester cleavage based on QM/MM computational studies.^[31] The prokaryotic homologs require the assistance of an amino acid to compensate for the role of the second metal, which correlates their cleavage mechanism with that of the single-metal-dependent nucleases. Specifically, the histidine (H214 in TmEndoV) in the catalytic tetrad (DDEH) acts as a catalytic base to deprotonate a nearby water molecule to favor nucleophilic attack on the phosphodiester bond. The metal-coordinated water molecule stabilizes the transition state and a nearby lysine residue (K139) then acts as a general acid catalyst for incision through leaving group activation (Figure 3A).

Figure 3:

Proposed catalytic cleavage mechanisms. A) One-metal mediated phosphodiester cleavage by prokaryotic EndoV. B) Two-metal mediated phosphodiester cleavage by hEndoV.

For eukaryotic EndoV, the same group computationally determined that human EndoV can catalyze the hydrolysis reaction through a single metal ion bound directly to the scissile phosphate group, providing an enhanced charge stabilization of the substrate.^[30] In a second pathway involving two metal ions, one of the magnesium ions (Mg_A²⁺) coordinates with the nucleophilic water molecule, the aspartate, and the scissile phosphate while the aspartate coordinates another nearby water molecule to activate the nucleophilic water molecule. As the nucleophilic attack in the first step proceeds, the lysine stabilizes the transition state intermediate through hydrogen bonding. Finally, while the lysine reverts back, the oxygen of the P-O bond is protonated by the metal (Mg_B²⁺) coordinated water molecule to complete the cleavage (Figure 3B). As highlighted in the section above, the in crystallo RNA hydrolysis experiments by Yang et al^[22] revealed yet another mechanism involving a third metal ion in mouse EndoV. Unlike in previous cases, the third metal ion directly interacts with substrate RNA, engaging the nucleophilic water molecule to coordinate the hydrolysis reaction and showcasing the feasibility of substrate assisted catalysis in EndoV. All of these mechanisms appear consistent with observations of EndoV and other nucleases, both structurally and biochemically, indicating that perhaps all three represent viable cleavage mechanisms, though the two-metal mechanism appears more favorable than the single-metal mechanism in simulation of human EndoV (the three-metal model was not tested computationally).

2.5. EndoV Evolution

Despite being dedicated DNA repair enzymes directed against deoxyinosine, the prokaryotic EndoVs as a whole exhibit promiscuity for inosine containing RNA substrates, making them Inosine-specific nucleases instead of restricting their activity to either DNA or RNA.^[27] In the case of eukaryotic EndoVs, it is hypothesized that function was further refined in response to increasing complexity, the compartmentalization of cellular machinery, and the advent of enzymatic inosine generation by ADARs. Together, these evolutionary steps are thought to have driven co-evolution of EndoV for careful modulation of inosine containing RNA.^[32] On par, the in vitro biochemical experiments revealed a drastic reduction in ability of eukaryotic EndoV to cleave inosines in DNA.^[25]

Understanding the RNA/DNA recognition bias of these Endonuclease V enzymes is crucial to exploring their native functions. Yan et al. provided structural insights for the evolution of RNA preference based in part on the similarities between EndoV and the RNase H1 and Argonaute proteins, which bind to RNA-DNA duplexes, yet only cleave the RNA strand of the duplex.^[22] Notably, RNase H1 and Argonaute are the only other members of this family besides EndoV which contain an active site glutamate residue which coordinates with the 2’-hydroxyl of the ribose sugar. This does not preclude the binding of DNA, though it does enforce a preference for RNA (or at least ribose at this position), as indicated by the increased cleavage rates of ssDNA having a ribose substitution at the cleavage site relative to unmodified ssDNA. One of the main structural differences between prokaryotic and eukaryotic EndoVs was actually found to be extended interactions with the downstream nucleic acid backbone outside the active site. In particular, insertions in the loops L₅₆ and L₇₈ allow eukaryotic EndoV to form extended interactions with the phospho-sugar backbone, with the structure of the loops being such that these interactions discriminate between DNA and RNA backbones. The importance of these loops in imparting RNA specificity was further validated by the loop swap experiments between TmEndoV and Ciona intestinalis EndoV (CiEndoV), resulting in up to a 40-fold preference for RNA over DNA by the hybrid TmEndoV.

3. Role of hEndoV in Cells

Human Endonuclease V is an inosine-specific ribonuclease expressed ubiquitously in human cells. Unlike its prokaryotic homologs which function as DNA repair proteins, hEndoV acts selectively on inosine-containing RNA transcripts and recognizes inosine residues generated as a result of adenosine-to-inosine (A-to-I) editing carried out by ADAR enzymes.^[32,33],[34] Alternative splicing of the ENDOV transcript leads to numerous splice variants, but only three isoforms, hEndoV-282, -308, and -309, encode full-length proteins having the conserved catalytic core for inosine recognition.^[35] Among these, hEndoV-309 is the most abundant and stable splice variant mRNA as well as the only protein isoform expressed in mammalian cells, as quantified by proteomic analyses. The relative abundance of hEndoV isoforms differs amongst primary and immortalized cells, suggesting that expression of the ENDOV gene and splicing of its transcript are influenced by cell type and proliferation state.

The localization of hEndoV is primarily cytoplasmic in human cells, which is consistent with the localization of most A-to-I edited RNAs. However, hEndoV can also be found in the nucleoli, which highlights its potential for additional roles in ribosomal RNA processing or surveillance.^[34] Under normal conditions, the enzyme prospectively monitors inosine-containing RNAs and helps preserve the integrity of the transcriptome. Conversely, stress conditions influence hEndoV localization. For instance, upon oxidative or translational stress it partially relocalizes to cytoplasmic stress granules along with many other proteins, including poly(A)-binding protein C1 (PABPC1), Tudor-Staphylococcal Nuclease (Tudor-SN), and ADAR1p150.^[36] While no direct interaction was found between PABPC1 and hEndoV, a pulldown experiment with overexpressed FLAG-hEndoV showed a potential RNA-dependent interaction that was abolished upon RNAse treatment. Incubation of hEndoV and excess PABPC1 was found to enhance the apparent binding affinity of hEndoV for inosine-containing RNAs in an electrophoretic mobility shift assay as well as increase the extent of cleavage of an IIUI-containing RNA substrate. Even though the same study has shown that physiological concentrations of various NTPs inhibit hEndoV cleavage activity, the ability of these NTPs to chelate the magnesium ions required for hEndoV cleavage in the in vitro assay put forth as evidence cannot be ruled out.^[36]

One key example of the cellular function of hEndoV can be found in the degradation of edited Alu dsRNA structures.^[37] Ribonuclease activity was reported that specifically cleaves both strands of double-stranded RNA containing multiple IU base pairs, such as the IIUI motif, which are found in extensively edited Alu regions of transcripts.^[38] Subsequent work identified hEndoV as the enzyme responsible for this cleavage activity^[39] and the activity was shown to be enhanced by Tudor-Staphylococcal Nuclease (Tudor-SN), which acts as a cofactor.^[40] Collectively, hEndoV and Tudor-SN may control turnover of extensively edited dsRNA structures, thus regulating the expression of genes that contain inverted Alu repeats.^[37] As hyper-edited RNAs are measurable in steady-state mRNA pools, their degradation is not persistent and is may be regulated by their subcellular localization. Interestingly, hEndoV and Tudor-SN both accumulate in stress granules during cellular stress.^{[36],[41],[42]} The IFN-inducible isoform ADAR1p150 also binds to edited dsRNAs through its Zα domain and colocalizes with Tudor-SN in stress granules.^[42],[43] These findings suggest that under cellular stress or viral infection, hEndoV, Tudor-SN, and ADAR1 may act together to hyperedit and subsequently remove or remodel hyper-edited RNAs, preventing their accumulation and the subsequent activation of innate immune sensors such as MDA5 or RIG-I.^[44]

Beyond RNA metabolism, hEndoV has also been linked to disease processes. In a diethylnitrosamine-induced hepatocellular carcinoma (HCC) model, EndoV knockout mice developed fewer and smaller tumors than wild-type animals.^[45] Loss of EndoV did not alter global RNA editing or ADAR expression but changed tRNA-derived fragment profiles, implying functions in RNA processing or stress signaling rather than direct regulation of editing. In vascular smooth-muscle cells, hEndoV promotes cell migration and atherosclerotic lesion formation, possibly through inosine-mediated stabilization of mRNAs.^[45] While still preliminary, these findings connect hEndoV to cellular stress, cancer, and vascular disease, suggesting that its physiological role extends beyond RNA surveillance to the control of stress-responsive gene expression. Taken together, hEndoV functions as a stress-responsive inosine-containing RNA sensor that modifies its localization and activity in response to the cellular environment. The association between hEndoV, RNA editing, and disease pathology suggest that it plays a broader role in RNA homeostasis and stress adaptation in mammalian cells, opening the door to exciting future studies to further understand the cellular function of EndoV.

4. EndoV Technologies for Inosine Detection

Dysregulated A-to-I editing has been linked to several diseases, such as cancer, autoimmune disorders, and neurodegeneration.^[46–48] This creates an inevitable demand for approaches to map and quantify editing across the transcriptome in vitro and in cells and tissues. Over the past decade, researchers have attempted to identify a selective reaction partner for inosine that could enable identification of A-to-I editing sites or enrichment of inosine-containing RNA from the total RNA pool.^[49–51] The first of these approaches utilized the reactivity of acrylonitrile with inosine through Michael addition to develop a technique called ICE-seq.^[49] The resulting adduct, N1-cyanoethylinosine, stalls the reverse transcription of RNA, selectively erasing the inosine-containing transcripts and allowing a comparative identification of editing sites after sequencing.^[49] While this method aids in the identification of editing sites, the lack of a functional handle on acrylonitrile makes enrichment of the edited transcripts difficult. To address this challenge, our lab employed acrylamidofluorescein, which can similarly react with inosine-containing RNA and has a fluorescein handle that can be used for immunoprecipitation, resulting in a 7-fold enrichment of inosine-containing transcripts (Figure 4A).^[50] This method was further improved by replacing the acrylamidofluorescein with N-(4-ethynylphenyl) acrylamide (EPhAA) to allow for post-labeling probe attachment using an azide-alkyne click reaction (Figure 4B).^[51] This modified approach provided better reagent solubility and up to 3-fold increase in labeling compared to acrylamidofluorescein, as well as 60-fold selectivity for inosine-containing RNA.^[51] Despite these advances, one universal challenge for these Michael acceptor reagents is their off-target reactivity with pseudouridine, which limits their use for detecting inosine modifications in biological samples.^[49–51]

Figure 4:

Chemical labeling approaches: A) Acrylamidofluorescein-mediated inosine-RNA labeling and enrichment B) N-(4-ethynylphenyl) acrylamide (EPhAA) mediated labeling followed by click reaction with Cy5 azide.

An alternative approach to inosine detection involves leveraging the molecular recognition capabilities of naturally occurring proteins and enzymes. In particular, the reader proteins that naturally recognize and bind to posttranscriptional modifications have been successfully utilized to capture and enrich the modified RNAs in techniques such as RNA immunoprecipitation (RIP) and cross-linking and immunoprecipitation (CLIP).^[52] The first extension of this approach for enriching inosine-containing RNAs utilized RNase T1, which is known to cleave RNA at the 3’ side of guanosine. By blocking guanosine (but not inosine) with glyoxal and borate, the selective cleavage was shifted to inosine. When using RNAs that are biotinylated at the 3’ end prior to RNase T1 incubation, streptavidin pulldown can be used to separate out the non-edited sequences.^[53,54] In 2013, Cattenoz et al. translated this protocol into iSeq for transcriptome-wide detection of A-to-I edited sites using high-throughput Illumina sequencing.^[55] Encouragingly, iSeq successfully captured 665 A-to-I edited sites with high confidence, but hyper-edited sites with clustered inosines could be potentially lost due to cleavage of RNA into fragments that are too short for sequencing.

In 2020, our lab demonstrated an alternative approach to repurposing naturally occurring enzymes by using EndoV as a selective molecular recognition partner for detection of inosine in vitro. In particular, we found that replacing the obligate Mg²⁺ cofactor with Ca²⁺ resulted in retention of the selective binding capability with inosine-containing nucleic acids, but eliminated the catalytic function, enabling EndoV to function as an ‘anti-inosine antibody.’ We first utilized this affinity reagent strategy to create Endonuclease V inosine precipitation enrichment sequencing (EndoVIPER-seq), for capturing and enriching edited transcripts in order to better detect low frequency editing sites (Figure 5A).^[56,57] As part of the optimization of EndoVIPER-seq, we tested binding of eEndoV to synthetic single-stranded RNA mimicking unedited (ssRNA-A) and edited (ssRNA-I) sequences, and we observed a 350-fold binding selectivity for ssRNA-I over ssRNA-A, providing insight into the binding preferences of eEndoV. Elaboration to dsRNA structures revealed that eEndoV binding is significantly diminished in this context, demonstrating that eEndoV has a preference for single-stranded nucleic acids when operating in binding mode. Fortuitously, we were able to overcome the low affinity for dsRNA through glyoxal-mediated denaturation and went on to use EndoVIPER-seq to uncover 27,000 novel editing sites in brain cells, with a 38-fold enrichment in read coverage across all tested samples.^[57]

Figure 5:

Leveraging selective recognition of inosine by EndoV for the development of assays to enrich, quantify, and image A-to-I editing in RNA, cell, and tissue samples. A) EndoVIPER, B) EndoVLISA, C) EndoVIA, D) EndoVIA 2.0.

In 2022, Chen et al. used the catalytic activity of hEndoV to create a cleavage-based assay, hEndoV-seq, that functions analogously to the RNase T1 assay described above but leverages the inherent selectivity of EndoV for cleavage at inosine.^[58] While able to achieve single-nucleotide resolution, throughput is somewhat limited by the use of Sanger sequencing in the workflow. To address this issue, Wei et al. developed a high-throughput hEndoV-based cleavage assay, Slic-seq, in which the RNA is 3’ oxidized and thus adaptor ligation can be selectively performed on cleaved sequences, which are then subjected to DNA library preparation for Illumina sequencing.^[59] This method claims a higher overlap of identified edited sites with the publicly available databases compared to EndoVIPER-seq.^[59] However, similar to RNase T1 approaches, the use of cleavage may limit detection of inosines that are in close proximity within a sequence.

While sequencing offers a convenient method for mapping A-to-I editing across the transcriptome, it can be costly and require long waiting times, and it is not amenable to high levels of multiplexing. Our lab recognized that with the availability of EndoV as an ‘anti-inosine antibody,’ we could develop assays that would enable rapid and high-throughput analysis of global editing levels. The first assay developed was EndoV-linked immunosorbency assay (EndoVLISA), which functions analogously to an ELISA to quantify global A-to-I editing levels in cellular samples (Figure 5B). Specifically, the RNA is biotinylated and glyoxylated and then bound to streptavidin-coated plates. The bound RNA is successively incubated with eEndoV-MBP, primary anti-MBP antibody, and secondary antibody conjugated to horseradish peroxidase (HRP) reporter enzyme. The HRP produces a chemiluminescent signal that can be directly correlated to the level of editing in the cellular RNA sample. Using this assay, we were able to detect both tissue-specific and disease-relevant changes in global editing levels, and to do so at a fraction of the time and cost of sequencing.^[60]

EndoVLISA allows quantitative detection of editing levels in cellular samples, but it does sacrifice information about cell-to-cell heterogeneity and the spatial distribution of edited RNAs in cells, both of which are crucial for studying the underlying link between editing and cellular function. Elaborating on our other technologies that harness the binding capability of eEndoV, we developed Endonuclease V immunostaining assay (EndoVIA) in order to directly image and quantify edited RNAs in cell samples (Figure 5C). Similar to the EndoVIPER workflow, the cells are treated with glyoxal to denature RNA structure, then primary and secondary antibodies are successively added to enable imaging of the edited RNAs. Using this method, we were able to quantify the change in editing between healthy and cancerous cells and perform single-molecule imaging to identify the subcellular localization patterns for edited RNAs across different cell types.^[61] As we sought to extend this technology further for use in tissue samples and organoids, a key challenge we recognized was the selectivity of eEndoV for single-stranded RNA. While glyoxal treatment could be used in cell culture to disrupt dsRNA structures, the highly crosslinked nature of preserved tissues would make this approach unfeasible. Fortuitously, we found that hEndoV was able to bind dsRNA in the cellular context, and we recently reported EndoVIA 2.0, which is capable of imaging and quantifying A-to-I editing in a wide range of samples, including fresh frozen tissues, FFPE tissues, and organoids (Figure 5D).^[62] Together, these assays demonstrate that the unique recognition capabilities of the individual EndoV orthologs can be harnessed to develop improved technologies for detecting inosine-containing RNAs in a range of contexts. We envision that these assays will advance the study of editing across multiple disease states and accelerate the discovery of new small molecule therapeutics capable of modulating A-to-I editing.

While natural A-to-I editing in human cells is only known to occur in RNA, engineered base editors have shown exciting promise for installing A-to-I edits in DNA as a form of gene therapy. Given the permanent nature of DNA edits, mapping and quantifying off-target editing is crucial for these technologies. To address this need, Liang et al. developed an EndoV-seq approach using TmEndoV. This method relies on Cas9 and TmEndoV working together to induce double-strand breaks in the DNA near the sites of A-to-I editing. Specifically, the Cas9 induces a nick on the opposite strand during the base editing process and the in vitro deaminated target genes or the extracted edited genomic DNA can be incubated with EndoV to cut the edited strand, resulting in a double strand break. This offers a convenient method for alignment to the reference genome for mapping of editing sites through whole genome sequencing. Even though the method is limited by the necessity of Cas9-mediated cleavage on the opposite strand, it stands superior over conventional digenome-seq to identify both on-target and off-target editing events.^[63] Thus, while inosine is rare in DNA, the ability of some EndoV orthologs to cleave near inosine sites in both RNA and DNA has proven to be of significant utility in biotechnology.

5. Conclusion

The discovery of EndoV from E. coli marked a significant milestone in establishing Nature’s ability to recognize inosine in nucleic acids.^[64] Later studies elucidated the intriguing evolutionary journey from the use of EndoV in prokaryotes for DNA repair to the ability of EndoV in eukaryotes to modulate epitranscriptomic modifications.^[65] While the sequences and structures of EndoV orthologs continue to be elucidated, many questions pertaining to the specific purpose and activity of EndoV in human cells remain unanswered. The preferential activity of hEndoV on hyper-edited transcripts resembling ADAR1-edited Alu element regions suggests a role in regulating the dsRNA response in cells, and has implications for cancers in which ADAR1 is overexpressed. In addition, even though preliminary in vitro studies indicate an enhanced activity of hEndoV in the presence of other enzymes such as PABPC1, their specific interaction mechanisms and in vivo function remain unexplored. Moreover, even though hEndoV 309 is the biologically abundant isoform, it stands as less studied compared to other isoforms, even for in vitro studies. Thus, understanding potential differences in activity between hEndoV isoforms represents an important area that would benefit from greater attention of the scientific field.^[66]

In parallel with understanding the natural biological roles of EndoV, its activity has been leveraged by a number of research labs to advance technologies for detecting inosine-containing RNAs. While some of these methods capitalize on the catalytic activity of EndoV, others intentionally eliminate this activity in order to allow EndoV to act as an inosine-specific affinity reagent. Together, this suite of tools is advancing the study of RNA editing in exciting directions, and in turn, the development of these tools is providing new insights into the substrate preferences and activity of EndoV.

Biographies

Jennifer M. Heemstra is the Chair and Charles Allen Thomas Professor of Chemistry at Washington University in St. Louis. She earned her B.S. in Chemistry from University of California, Irvine and Ph.D. in Chemistry from University of Illinois, Urbana-Champaign. She has long been fascinated with the field of supramolecular chemistry, and together with her research group, she enjoys harnessing the exquisite molecular recognition properties of biomolecules for applications in biomedicine and related fields.

Prasanth Thota received his M.Sc in Organic Chemistry from Sri. Y. N. College, Narsapur and worked on Enzyme Catalysis as a project assistant in Saravanan Research Group at University of Hyderabad. He joined the Heemstra Lab, Washington University in St. Louis, in 2023 for his PhD. His research focuses on developing a high-throughput compatible assay to track A-to-I editing and exploring the specificity of the human Endonuclease V enzyme.

Tyson Todd is currently a postdoctoral researcher at Washington University in St. Louis under the supervision of Professor Jennifer Heemstra. He completed his Ph.D. at Washington University in St. Louis under the supervision of Professor Kendall Blumer. His research focuses on utilizing directed evolution to develop novel protein-based probes of post-transcriptional modifications with applications in understanding the prevalence, localization, and phenotypic effect of specific post-transcriptional modifications in normal and disease states in human cells.

Devanshi Purohit is a Ph.D. candidate in Chemistry at Washington University in St. Louis in Prof. Jennifer Heemstra’s lab. Her research focuses on developing tools for high-throughput screening of small-molecule A-to-I RNA editing modulators in breast cancer cell models. She integrates cancer biology, molecular imaging, and RNA biochemistry to uncover how A-to-I editing influences disease pathology in breast cancer, with the broader goal of enabling precision therapeutics.

Weina Cheng is a Ph.D. student at Washington University in St. Louis, conducting her research under the supervision of Professor Jen Heemstra. She earned her bachelor’s degree from the University of Michigan. Her current research focuses on protein engineering of RNA-binding proteins and RNA post-transcriptional modifications.

Aastha is currently a Ph.D. student at Washington University in St. Louis under the supervision of Professor Jennifer Heemstra. She completed her BS-MS dual degree from the Indian Institute of Science Education and Research, Mohali from India. Her research focuses on exploring the changes in subcellular localization of A-to-I edited RNA and editing levels across the transcriptome in different neurodegenerative diseases.