Recommendations for detection, validation, and evaluation of RNA editing events in cardiovascular and neurological/neurodegenerative diseases

RNA editing, a common and potentially highly functional form of RNA modification, encompasses two different RNA modifications, namely adenosine to inosine (A-to-I) and cytidine to uridine (C-to-U) editing. As inosines are interpreted as guanosines by the cellular machinery, both A-to-I and C-to-U editing change the nucleotide sequence of the RNA. Editing events in coding sequences have the potential to change the amino acid sequence of proteins, whereas editing events in noncoding RNAs can, for example, affect microRNA target binding. With advancing RNA sequencing technology, more RNA editing events are being discovered, studied, and reported. However, RNA editing events are still often overlooked or discarded as sequence read quality defects. With this position paper, we aim to provide guidelines and recommendations for the detection, validation, and follow-up experiments to study RNA editing, taking examples from the fields of cardiovascular and brain disease. We discuss all steps, from sample collection, storage, and preparation, to different strategies for RNA sequencing and editing-sensitive data analysis strategies, to validation and follow-up experiments, as well as potential pitfalls and gaps in the available technologies. This paper may be used as an experimental guideline for RNA editing studies in any disease context.


INTRODUCTION
RNA editing is a post-transcriptional modification that alters the sequence of an RNA transcript.It was first discovered by Benne et al.  in a mitochondrion-encoded mRNA in Trypanosoma brucei. 1 Wagner et al. showed that RNA editing also occurs in mammalian cells. 23][4] A-to-I, which is the most common form of RNA editing, is mediated by the ADAR (adenosine deaminases acting on RNA) family of enzymes, ADAR1, ADAR2, and ADAR3. 5,6ADAR1 and ADAR2 are catalytically active.ADAR1 is ubiquitously expressed while ADAR2 expression is very low in some tissues.ADAR3 has no proven catalytic activity, and its expression is limited to the brain. 7ADAR3 has been shown to inhibit ADAR1-induced RNA editing. 8ADAR1 and ADAR2, but not ADAR3, deaminate adenosines to inosines, which are interpreted as guanosines by the translational and splicing machinery. 5ADARs likely have non-editing-related functions too, but these are not discussed here.
The deamination of cytosine to uridine is catalyzed by the APOBEC (apolipoprotein B mRNA editing enzyme catalytic polypeptide-like) enzyme family.In mammals, APOBEC1 is the main C-to-U enzyme and, differently from ADARs, it needs an essential co-factor (A1CF or RBM47) and auxiliary proteins for deamination.This protein complex, called the editosome, targets single-stranded RNAs by recognizing an 11-nucleotide AU-rich mooring sequence downstream of the editing site. 9A editing events may affect RNA localization, structure, stability, and transcript processing of both coding and noncoding RNAs (ncRNAs).10 Editing is crucial for cell and tissue homeostasis and a variety of human diseases, including both cardiovascular and neurological disorders, have been linked to its deregulation.[11][12][13][14][15] Editing in protein-coding mRNA sequences can lead to non-synonymous changes resulting in amino acid alterations (protein recoding or stop codon introduction) and novel protein isoforms.Although this event is not frequent, it takes place in important mammalian genes, such as those encoding the potassium channel Kv11, 16 the glutamate receptor subunit GluR2, 17 and the a3 subunit of GABAA (g-aminobutyric acid type A) receptor (GABRA3) 18 and Filamin A. 19 Most RNA editing events, however, occur in 5 0 and 3 0 untranslated regions (UTRs) and introns of protein coding genes that harbor Alu repeats.11,[20][21][22] Editing in UTRs may affect gene expression through nuclear retention, RNA degradation, and translational regulation, while editing in intronic regions could influence alternative splicing of edited transcripts by modifying or eliminating splice donor and acceptor sites.23 Editing also affects ncRNAs.25 Moreover, editing in mature microRNA seed sequences can influence the recognition of binding sites in target mRNAs, thereby changing the microRNA's set of target mRNAs, its "targetome."12,[26][27][28] In addition, RNA editing has the potential to modify microRNA target binding sites, with significant regulatory and functional implications.22,29 Long non-coding RNA (lncRNA) editing can lead to their nuclear retention and alter their biological function by disrupting their binding sites for DNA and RNA molecules as well as for RNA binding proteins.30 Biogenesis of circular RNAs (circR-NAs) can also be influenced by editing of the parent sequence. 31 Thregulation of RNA editing events is a complex process, as demonstrated by studies highlighting different regulatory mechanisms.For instance, several studies elucidated the influence of genetic loci in regulating RNA editing events.32,33 In addition, other investigations have shown that RNA binding proteins can play a role in modulating RNA editing.34 Furthermore, the majority of crucial editing regulators exert their influence in a cell-type-specific fashion.26 These findings underscore the intricate nature of RNA editing regulation and the involvement of various factors in shaping the editing landscape.
RNA editing is clearly a widespread and important mechanism of gene regulation in both health and disease.However, studying RNA editing in biological and clinical settings has its challenges.Therefore, we here provide an experimental guideline for RNA editing studies, using studies from the cardiovascular and neurovascular/neurodegenerative fields as an example, but with applicability to all biomedical research fields.An experimental pipeline, from sample collection to functional and clinical validation, is presented in a single comprehensive figure (Figure 1).

Sample collection, storage, and preparation
Sample collection, storage, and preparation are of the utmost importance for any RNA sequencing (RNA-seq) experiment, as they dictate its outcome.The protocol for optimal sample collection, independent of RNA editing, is an overall standard procedure but, depending on the type of the sample, there are some differences.

Plasma
Biofluids are intensively studied when looking for new RNA-based biomarkers of various diseases.Among these, plasma, although a great candidate for this as it is readily accessible, is somewhat tricky (Figure 2).One crucial aspect that should be kept in mind when collecting plasma for RNA studies is that heparin should not be used as an anti-coagulant, as it inhibits further PCR assays. 35It is recommended that EDTA or citrate tubes be used for plasma collection.If this is not available, heparinase treatment of the RNA isolated from heparin plasma samples is needed before further analysis. 36Furthermore, the use of hemolyzed plasma should be avoided for microRNA analysis, as hemolysis can alter the plasma levels of microRNAs that are enriched in red blood cells, which can lead to inaccuracies and misinterpretation of the final results. 37In addition, one should consider the use of platelet-poor vs. platelet-rich plasma and avoid unintended platelet activation as platelets carry a large RNA load, including many microRNAs. 38

Cell cultures and tissue samples
The handling of cultured cells or tissues with a view to isolate RNA usually entails the lysis of cells with the help of a cell scraper (for cultured adherent cells) or a tissue homogenizer (for tissues) using either solvents such as TRIzol (Waltham, MA) or other lysis buffers supplied by commercially available kits for RNA isolation.It is important to note that, until recently, due to problems with RNA degradation and formaldehyde-mediated RNA modifications, paraformaldehyde-fixed cells or tissues were not considered as a suitable matrix when RNA-based molecular studies were involved. 39This issue has been overcome with a recent high-throughput RNA-seq technique called fixed droplet RNA-seq that was shown to preserve RNA integrity and transcriptomic information in PFA-fixed and permeabilized cell cultures. 40 is noteworthy that there are some commercially available solvents, such as DNA/RNA Shield (Zymo Research, Irvine, CA) that can be added to biological samples upon collection to preserve and protect nucleic acid integrity.It is important to consider that the addition of such solvents further dilutes the sample and it might not be suitable for biological samples with a low RNA yield, such as plasma.Cell culture supernatant should be treated in a similar fashion as plasma when it comes to both storage and RNA isolation, given the fact that RNA concentrations are low and extracellular vesicles (EVs) are often present.

Sample storage
Storage conditions (time and temperature) of the samples (tissues, cells, or biofluids) during the pre-analytical phase should be strictly controlled to minimize technical variation.Some subtypes of ncRNAs, e.g., microRNAs and circRNAs, are highly stable in different types of samples due to their resistance to degradation. 41evertheless, since RNA is generally highly unstable, samples should be processed as quickly as possible.When immediate specimen preparation is not feasible, it is recommended to store samples at 4 C or on ice. 42For long-term storage, a temperature of À80 C is recommended over À20 C to optimize RNA integrity.The use of commercially available RNA stabilization reagents constitutes an interesting option to preserve RNA during storage and thawing.However, for studies focusing on EVs, the stability of EVs remains an issue and special RNA stabilization reagents are needed.For example, Görgens et al. have shown the use of PBS supplemented with human albumin and trehalose to enhance the stability and improve long-term storage of EV-containing samples at À80 C. 43 Consistency in the storage conditions is imperative during sample processing, not only to avoid RNA degradation but also to improve the reproducibility of the findings.Consequently, the rigorous documentation of storage information and the inclusion of these data in scientific publications is fundamental.

RNA isolation and storage
Obtaining high-quality RNA is also one of the most critical steps for further accurate RNA-seq results.There are three basic RNA isolation approaches with different characteristics and the most suitable one depends on the sample material, the quantity and the type(s) of RNA that needs to be acquired.Organic extraction techniques are widely regarded as the gold standard and they rely on the different solubility of cellular components in solvents such as phenol, ethanol, or isopropanol. 44,45This methodology is ideal for samples high in nucleases (e.g., pancreas) or lipid content (e.g., brain or adipose tissue).Filter-based spin column RNA extraction techniques are based on the absorption of RNA to specific surfaces in the presence of chaotropic salts and is the easiest and safest method available for high-throughput sample processing needs. 46Finally, magnetic particle RNA extraction is an easy-to-automate alternative which enables high-throughput procedures. 47Until further processing, it is important that RNA be stored in an RNase-free environment, preferably at a temperature of À80 C.

Different RNA-seq approaches General transcriptome sequencing
Depending on the type of RNAs of interest (coding or noncoding; longer or short), there are some required steps that should be followed when preparing sequencing libraries.
One approach that researchers often use is the selection or enrichment of poly(A) transcripts.In eukaryotes, most mRNAs and many lncRNAs contain a poly(A) tail, 48,49 which can be used as a technical tool for selection, either with magnetic or cellulose beads coated with oligo-dT molecules or with oligo-dT priming for reverse transcription (RT). 50 alternative approach is the depletion of ribosomal RNA (rRNA), as it is the most abundant RNA type in all cells and it is often of limited interest to most studies.To deplete rRNAs, the one widely used method is the hybridization of rRNA-specific probes, followed by depletion with streptavidin beads. 51More specific to ncRNA experiments, the two most widely used methods for rRNA depletion are probe-directed degradation (rRNA targeting by anti-sense DNA oligos and subsequent digestion by RNase H) 52 and not-so-random priming. 53It should be noted that ribodepletion still allows further poly(A) selection of transcripts, so the choice of the technique used should be based on the question or the aim of the conducted research.
The next step consists of sample fragmentation before undergoing RT, which is achieved either by divalent cation treatment with specific enzymes or by using alkaline solutions. 54To retain the information pertaining to strand origin, the standard RNA-seq protocol should be modified as follows: during cDNA synthesis, the second-strand synthesis continues as normal except the nucleotide mix includes dUTPs instead of dTTPs; then, after library preparation, a secondstrand digestion step is added, which ensures that only the first strand survives the subsequent PCR amplification step and hence the strand information of the libraries. 55ternative approaches RNA-seq alternatives following different approaches should be used depending on the average size of RNAs of interest.Whole-transcriptome approaches sequence both protein coding RNAs (mRNAs) and lncRNAs, due to their structural similarities. 56It is most suitable for the discovery of novel transcripts, but some shortcomings are present due to the requirements of large amounts of RNA input and a certain bias due to different sequencing length reads. 57Sequencing of circR-NAs follows the standard workflow of RNA-seq described above, with an additional step that consists of the digestion of linear RNAs beforehand. 58

Small RNAs
Small ncRNA sequencing, such as microRNA-seq, follows the same principle as lncRNAs, with some modifications.Due to their short size, microRNAs need an extension step either by ligation or by polyadenylation, which may introduce bias, 59 which can be mitigated by the use of unique molecular identifiers, 60 which help distinguish reads generated from an identical molecule that was amplified by PCR.

Targeted approaches
Targeted RNA-seq is a specific category that enables selection of specific set of transcripts from defined genes or regions of interest.It is a useful technique for low-expressed transcripts that cannot be easily analyzed through whole-transcriptome sequencing.Until now, the two general approaches commonly used are target capture [63][64][65][66][67] and amplicon sequencing. 68Target capture is based on the selection of Figure 3. Overview of the main approaches for small RNA-seq analysis (A) Original two-adaptor ligation protocol for small RNA-seq analysis.(B) Improved two-adaptor ligation methods for small RNA-seq analysis.The reduction of PCR or ligation bias can be accomplished through randomized adaptors in the ligation steps or unique molecular identifiers in the amplification step.(C) Ligation-free approach for small RNA-seq analysis through polyadenylation of the 3 0 end and template switching method.Created with BioRender.com.specific regions of interest through hybridization of RNA-seq libraries to a set of biotinylated probes.This method does not require the presence of a poly(A) tail, thus it is ideal for low abundant or degraded RNA.On the contrary, amplicon sequencing utilizes gene-specific primers for the amplification of cDNA targets in combination with a primer for poly(A) tail priming.Target capture seems to provide higher complexity and consistency, while amplicon sequencing has higher on-target rates. 68

mmPCR-seq
The mmPCR-seq method, a large-scale amplicon sequencing technique, has been demonstrated to detect A-to-I editing in both midlength and longer RNAs. 69The method utilizes RT to cDNA followed by amplification of multiple targets in the same pool and subsequent deep sequencing of the resulting products.While amplified sites have been 150-350 base pairs in length in previous studies, it is important to note that target RNAs can be much longer. 69,70The mmPCR-seq method detects RNA editing in the same way as other sequencing methods, through finding A-to-G mismatches to the genome; however, because of the pre-amplification step, mmPCR-seq has been confirmed to precisely measure RNA editing levels in samples even after preamplification of $1,000-fold of low-quantity samples. 69

Single-cell sequencing
Finally, single-cell RNA-seq (scRNA-seq) and single-nucleus RNAseq (snRNA-seq) are two powerful techniques that enable the delineation of transcriptomic cell-to-cell differences, revealing sub-populations with distinct molecular and functional characteristics. 71This is especially useful for studies concerning heterogeneous cell populations involved in the development of cardiovascular 72,73 and neurodegenerative diseases. 74,75RNA editing studies in mice revealed differences in the editing levels between neurons and glial cells, 76,77 while in Drosophila RNA editing levels vary across different neuronal types. 78Cell-type-specific RNA editing was also found in the neurons, glutamatergic neurons, medial ganglionic eminence-derived GABAergic neurons, and oligodendrocytes in the human brain. 79hallenges concerning single-cell analysis include low inputs of RNA for library construction, difficulty in the detection of low abundant transcripts, 80,81 and increased technical and biological noise due to cell-cycle differences. 82Since RNA editing analysis requires high sequencing coverage, the detection of RNA editing events in singlecell data is really challenging.Recently, a novel computational approach that attempts to overcome all the aforementioned technical difficulties has been developed. 83The novelty of this method is based on the establishment of a pseudo-bulk RNA-seq dataset for each cell type, by integrating the aligned reads of all cells of the same cell type, which significantly increases the sequencing depth and allows the utilization of existing bioinformatics tools for RNA editing detection.

RNA modifications
Despite the valuable information that short-read sequencing provides, it is often challenging to capture the true whole diversity and plasticity of all RNA modifications, due to amplification bias (e.g., non-sequence altering modifications are not included in the ampli-cons) and assembly difficulties. 84,85This is where long-read thirdgeneration sequencing technologies are significant alternatives.These platforms can potentially capture information about assorted modification types simultaneously, in full-length transcripts, at single-molecule resolution, and at a genome-and transcriptome-wide scale. 85At the present time, there are two major approaches: (1) single-molecule real-time (SMRT) sequencing technologies 86 and (2) nanopore-based sequencing technologies. 87SMRT sequencing can be used for the detection of several DNA modifications, such as 6-methyladenosine (6mA), 4-methylcytosine (4mC), 5mC, and 5-hydroxymethylcy tosine (5hmC).However, the detection of RNA editing and other RNA modifications with SMRT is not yet applicable due to technical restrictions.On the other hand, nanopore sequencing is able to identify both DNA and RNA modifications, not only in naturally occurring alterations but also in artificially induced ones. 85[91]

Various bioinformatic approaches to the data analysis
There are many challenges in the computational detection of RNA editing events from RNA-seq data, that still need to be addressed, such as the artifacts introduced by high rate of sequencing errors, genomic mutations, the accurate identification of de novo editing sites, the tissue-specific editing and the variance in RNA editing percentage in each different editing position.
RNA editing sites can either be detected using RNA-seq data or with a combination of RNA-seq and DNA sequencing (whole-genome sequencing) data from the same sample/individual, to achieve low false positive detection rates due to single-nucleotide polymorphisms (SNPs).When genomic data are unavailable the utilization of SNP annotation databases is crucial.For example, Srivastava et al. demonstrated that RNA editing sites can be predicted with high confidence in mouse brain by eliminating known mouse SNPs. 15A critical aspect of RNA editing detection analysis is the sequencing depth and coverage of RNA-seq input.3][94] The inclusion of both biological and technical replicates in the detection pipeline is also an important factor that can increase sensitivity of RNA editing detection. 93rious bioinformatic tools and algorithmic choices have been developed in the past few years and depending on the design of the RNAseq experiment, the available data, and the input parameters the most suitable choice may differ (Table 1).
REDItools 96 is the first published software for genome-wide RNA editing site detection and provides both A-to-I and C-to-U variant detection.It is suitable for both RNA-seq and DNA-seq data from the same sample/individual or RNA-seq data alone and uses pre-aligned reads in Binary Alignment/Map (BAM) format as input.It performs de novo and known editing site detection and utilizes a wide range of filters and quality control checks for the identification of false positives, especially near intronic splice sites, read-ends and homopolymeric regions.REDItools provides additional scripts for post-processing output tables.These scripts enable the filtering of candidate editing sites based on known annotations, as well as the identification of ambiguous alignments using the tool Blat (BLASTlike alignment tool).Identification of ambiguous alignments is critical to RNA editing site detection.
RNAEditor 97 provides comprehensive A-to-I variant detection and can be used both from a command line or a graphical user interface.It utilizes RNA-seq data and uses FASTQ files as input that are consequently subjected to a mapping step with BWA aligner. 98It can detect both de novo and known RNA editing sites and includes a clustering algorithm for "editing islands" detection, through a density-based spatial clustering of applications with noise (DBSCAN). 99S-Scanner 100 is a fast software package tool for wide identification of RNA editing sites that requires matching RNA-seq and DNA-seq data.FASTQ format or BAM files can be inserted as input for the analysis.RES-Scanner detects both de novo and known editing sites by implementing binomial statistical tests 101 on possible RNA editing sites, from which the user can adjust the threshold for true positives.RED-ML 102 is an efficient RNA editing prediction tool that employs machine learning approaches.Like other tools, RED-ML does not depend on known edited sites and can predict novel edited sites with high accuracy.The tool is designed to process BAM files as input for RNA detection analysis after raw reads alignment.In addition, RED-ML is incredibly fast, making it an ideal choice for researchers looking to predict RNA editing events.GIREMI 103 is a software tool for RNA editing prediction that uses RNA-seq data as input, without matching DNA-seq data require-ments.Alignment of raw reads is required prior to use, as it only accepts BAM files as input for RNA detection analysis.It can integrate biological replicates into one dataset for higher accuracy or RNA editing detection.GIREMI's algorithm is based on mutual information between editing sites in RNA-seq data and SNPs, and thus, is suitable only for diploid genomes.Furthermore, a new function for RNA editing detection by long reads has been recently released, applied to PacBio RNA-seq data (L-GIREMI). 104SIC 95 is a tool for adenosine to inosine RNA editing sites identification and classification based on an alignment graph model and multiple filtering steps.This pipeline can be utilized for all organisms and is suitable for any number of RNA-seq datasets.RESIC enables the detection of RNA editing sites in both repetitive and non-repetitive regions, as well as identify hyper-edited sites.It can also optionally exclude polymorphism sites based on DNA and/or ADAR-mutant RNA-seq datasets.Although this tool provides many functionalities and options, it is not considered user friendly, as it requires manual manipulation of the provided Python script to specify the input data and cannot provide file names as command options.
JACUSA 105 detects all RNA nucleotide variants by comparing both RNA-DNA-seq and RNA-RNA-seq data (Figure 4).Unlike any other software tool, it can implement information from multiple experiments, such as biological replicates and different conditions for RNA-RNA differences detection.JACUSA can also integrate information from different library types, such as firstor second-strand cDNA libraries.Recently, an updated tool of JACUSA (JACUSA2) has been released, which presents better time performance and more complex read signatures. 106nally, SPRINT 107 is an SNP-free package toolkit for comprehensive A-to-I and C-to-U RNA editing sites detection (Figure 5).The novelty of this tool is that it can identify RNA editing sites (RES) without the

Editing databases
Databases that host collections of both predicted and validated editing sites are very useful for data comparison, filtering and validation.Currently, RNA editing events are mostly annotated in four main databases: EDK 116 (https://ngdc.cncb.ac.cn/edk/), REIA (http://bioinfo-sysu.com/ reia/), 117 REDIportal 118 (http://srv00.recas.ba.infn.it/atlas/) and TCEA (http://tcea.tmu.edu.tw). 119EDK is a manually curated database of RNA editing events in mRNAs, miRNAs, lncRNAs, viruses, and RNA editing enzymes, known from the literature to be associated with human diseases.TCEA and REIA integrate RNA editing events specifically focused on cancer research, while REIA also allows both editing profiling and interactive analyses with cancerrelated indices.REDIportal is the largest RNA editing resource containing approximately 16 million A-to-I events derived from 9,642 human RNA-seq samples and about 107,094 A-to-I mouse events from RNA-seq data.

In silico analyses of potential biological function
Rapidly progressing sequencing technology has resulted in an explosion of big data 120 and a subsequent increase in the detection of RNA editing events. 23Computational analyses provide mechanistic insights into the large amount of data produced in high-throughput studies.Depending on the different biological questions and produced data, there is a wide variety of approached methods, as well as databases to use.
RNA editing detection follows data assessment and RNA editing sites characterization and validation, all critical and necessary steps for potential connection with a molecular function.RNA editing events can result in codon changes, altered structure, stability, and transcript processing. 13,23Moreover, RNA editing sites can significantly impact the secondary structures of primary/precursor microRNAs and lncRNAs and their overall molecular interaction patterns. 1213][124][125][126][127][128][129] ANNOVAR is an open-source commandline software tool (http://www.openbioinformatics.org/annovar/) that uses pre-compiled annotation databases to annotate SNVs from diverse genomes.Variant effect predictor (VEP) (https://www.ensembl.org/Tools/VEP) is also a well-known and powerful software tool that can be used to analyze data from any species that has an assembled genome sequence and an annotated gene set. 130This tool includes a wide range of reference data, including regulatory regions, clinical significance information and biophysical consequences of variants.VEP is available as a command-line software, but also as a user-friendly online tool.Open Custom Ranked Analysis of Variants Toolkit (OpenCRAVAT) (https://opencravat.org/index.html) is a newly developed open-source software tool, combining a wide variety of diverse data resources and computational prediction methods.OpenCRAVAT can be utilized for the annotation of human genetic variation as well as for variant and gene prioritization. 129It is available both as a command-line and graphical user interface and is suitable only for human analysis.

Pathway enrichment
Pathway enrichment analysis is an important part of RNA editome studies as it provides potential mechanistic insights and biological interpretation of the data.2][133][134][135][136][137][138] Depending on the platform, there is a variety of structures, pathway content information and visual formats.To name a few, Metascape (https://metascape.org/gp/index.html#/main/step1) is a web-based portal design and provides a comprehensive and upto-date list of annotation information and analysis resources from 40 independent knowledgebases. 139EnrichR (https://maayanlab.cloud/Enrichr/) is also a popular and comprehensive gene set enrichment analysis web server, including more than 200 libraries and over 400,000 terms. 133,140It can integrate the knowledge from many wellregarded projects and provide synthesized information about genes and gene sets.It also offers visualization of the results through bar graphs, cluster-grams, scatterplots, volcano plots and other.

ncRNA interactions
A unique category of in silico analyses in RNA editome data includes functional analysis of microRNAs and lncRNAs based on their targets, as the majority of the RNA editing sites have unknown functions and exist in noncoding regions of the genome. 141There are several resources that store information about microRNAs, such as miRbase, which is one of the main databases that includes a complete micro-RNA catalogue, 142 DIANA-TarBase that contains experimentally verified microRNA to gene interactions, 143 DIANA-miRGen which is a database of microRNA transcriptional regulation, 144 miRCarta 145 and mirGeneDB. 146There are also several microRNA target prediction tools based on sequence complementarity.DIANA-microT-CDS is a web server hosting in silico predictions of microRNA-mRNA interactions. 147miRWalk is an open-source platform that can generate both predicted and validated miRNA-binding sites of known genes in human, mouse and other model organisms. 148IANA-miRPath is a tool that enables the functional annotation of a microRNA or the combined effect of multiple microRNAs, as well as the identification of microRNA-controlled pathways. 149IANA-mirExTra allows the identification of microRNAs that control mRNAs, transcription factors (TFs) regulating mRNAs and TFs regulating microRNAs between two conditions. 150TargetScan is an online prediction tool model that utilizes the presence of conserved 8mer, 7mer, and 6mer sites for the detection of biological targets of miRNAs. 151Another popular algorithm is TarPmiR, which enables the prediction of microRNA target sites using four different machine learning methods to CLASH (crossinking, ligation and sequencing of hybrids) data. 152Finally, there are also many databases that host information about lncRNAs.LNCipedia 153 and lncRNome 154 provide numerous human lncRNA entries including primary sequences and predicted secondary structures.DIANA-LncBase is the first database to include experimentally supported information about microRNA-lncRNA interactions. 155LNCediting is a user-friendly database providing customized tools to predict functional effects of novel editing. 121

Experimental validation of RNA editing events
As described above, new RNA editing events can be found using RNA-seq.However, the higher cost and complexity of processing samples and analyzing the data are significant disadvantages that can limit the practical use of RNA-seq in certain settings.Highthroughput sequencing, e.g., next-generation sequencing allows sequencing millions of RNA fragments simultaneously and it is used to detect RNA editing events genome-wide. 156Sequence data is then aligned to the reference genome.However, infrequently read sites may be incorrectly labeled as a misread or short reads may be aligned incorrectly. 156Validation of individual editing events by different methods is therefore necessary (Figure 6A).

RT-qPCR
RT-qPCR with custom primers is a widely used method to analyze relative expression of different RNAs. 157However, its ability to properly distinguish single-nucleotide changes must be confirmed for each individual editing event. 26,158Particularly for small RNAs such as mi-croRNAs a problem in distinguishing microRNA isoforms arises, because they often differ by only one nucleotide, and closely related isoforms may produce background signals in qPCR due to mismatches in primer annealing.Novel methods for validation of suspected RNA editing sites have been developed to overcome this problem, such as two-tailed RT-qPCR 157 and microfluidics-based multiplex PCR and deep sequencing (mmPCR-seq). 69o-tailed RT-qPCR can be used to reliably quantify A-to-I modifications of microRNAs. 157,158It utilizes two short, approximately 5-10 nucleotides in length, hemiprobes, connected by a hairpin structure.The target edited nucleotide is designed inside the short primer, and a single-nucleotide mismatch has been confirmed to disturb the pairing of the primer and the target RNA, making it possible to distinguish even single-nucleotide differences.

Southern blotting
An alternative method for detecting both A-to-I and C-to-U RNA editing in mid-length to long RNAs is analyzing cDNA fragments by Southern blotting/electrophoresis after digestion by sequence-specific restriction enzymes to reverse-transcribed and PCR-amplified cDNA. 27,159The objective of the electrophoresis method is cleaving the wild-type isoform using a specific restriction enzyme, while RNA editing prevents the enzyme cleaving the isoform, or vice versa.Importantly, synthetic cDNA is used to confirm both specificity and efficacy of the restriction enzyme, and genomic DNA is used to confirm the absence of an A-to-G SNP (van der Kwast et al. 27 ; supplemental information).Computer-assisted densitometric analysis can be used to objectively measure the levels of different isoforms. 27Alternatively, undigested amplified cDNA can also be analyzed by Sanger sequencing and compared with genomic DNA sequences. 27nctional validation and confirmation of clinical relevance of RNA editing RNA editing can affect both expression and function of all species of RNAs.Therefore, in this part of the article, we discuss methods to functionally validate as well as evaluate the clinical impact of RNA editing, using studies from the cardiovascular and neurological fields as examples.
Figure 6.Experimental validation and clinical relevance of RNA editing events (A) Various methods to discover, detect, validate, and quantify individual RNA editing sites.Top to bottom, the panel shows RNA-seq, allele-specific RT-qPCR, twotailed RT-qPCR, mmPCR sequencing, and restriction fragment-length polymorphism analysis of cDNA by southern blotting.(B) RT-qPCR to help in studying the impact of RNA editing on the expression of parent RNAs (mRNAs, lncRNAs, circRNAs, or miRNAs) or their downstream targets.Western blotting helps to study the effect of RNA editing on the protein expression of edited parent coding RNA or its downstream protein target/s.Luciferase reporter gene assays help in validate any alterations in the targetome of edited RNAs.(C) Methods to study the role of RNA editing enzymes.From top to bottom the panel shows RT-qPCR, western blotting, RNA binding protein immunoprecipitation (RIP), and various methods for DNA/RNA modulation.(D) Methods to study the functional effect of RNA editing events.CRISPR-based techniques such as RESTORE and RESCUE can be used to generate unedited and edited RNA molecules for in vitro or in vivo studies.Phenotypic assays such as proliferation assays, apoptosis assay, or cell migration assay help to identify the functional role of RNA editing in disease models.(E) Focus on clinical relevance.Clinical studies collecting patient and control blood samples or tissue biopsies can be used to verify association between RNA editing and disease or between RNA editing and drug (in-)sensitivity, for example.Furthermore, edited RNAs or editing enzymes may serve as novel disease-specific biomarkers.Created with BioRender.com.

Review
To functionally validate the impact of RNA editing, the effect on its parent RNA expression, its resulting protein (for mRNAs), and/or its downstream targets (for both coding and ncRNAs) must be analyzed 26,160 (Figure 6B).

RNA expression
Potential effects of editing on RNA expression can be measured by RT-qPCR assays designed for both protein-coding and ncRNAs.As an illustration, Paul et al. used qPCR to demonstrate that four distinct hypo-edited microRNAs were downregulated in brain samples of patients with glioblastoma multiforme (GBM) in contrast to corpus callosum samples obtained from non-GBM deceased patients, confirming effects of RNA editing on the expression of the mature microRNAs. 161

Protein expression
To determine modulations in protein expression caused by RNA editing of mRNAs, classical western blotting or immunohisto/cytochemistry can be employed, such as done by Jain et al., who looked at the effects of an RNA editing-directed amino acid change on the protein expression of Filamin A using western blotting.The authors showed that the amino acid change (Q-to-R), caused by a highly conserved A-to-I RNA editing event, can lead to a decrease of Filamin A protein expression in primary mouse lung fibroblasts. 162 validate indirect effects of RNA editing on gene expression, luciferase reporter gene assays are commonly used.For example, increased editing of a specific subset of microRNAs under ischemia alters the microRNAs' targetomes, as demonstrated by van der Kwast et al., who cloned target mRNA 3 0 UTR sequences with single or multiple microRNA binding sites (unedited vs. edited) into the PsiCheck2 luciferase reporter gene vector.Co-transfection of the vector with either unedited or edited microRNA mimics in HeLa cells resulted in significant changes in luciferase activity, confirming effects on target binding and gene expression. 26,27

ADARs and APOBECs
The genes encoding the ADAR and APOBEC families of deaminases can produce several variants or isoforms of these enzymes.Thus, it is important to characterize the enzymes and determine which family member(s) and isoform(s) are responsible for the observed RNA editing patterns (Figure 6C).To accomplish this, RT-qPCR at mRNA level and western blotting or immunohisto/cytochemistry at protein level can detect changes in expression of these enzymes. 26,27,160,161,163,164In a study by Kokot et al., investigating failing human hearts, both ADAR1 and ADAR2 were examined.While qPCR did not show a regulation at transcriptional level, western blotting showed that both ADAR1 and ADAR2 were regulated in heart failure samples compared with samples from non-failing hearts. 163sides enzyme expression, direct interaction of the deaminases with the edited RNA should be analyzed.An effective way to achieve this is by using RNA immunoprecipitation (RIP) with an antibody specific to the deaminase in question.Many protocols relating to RIP to iden-tify targets for RNA-binding proteins (RBPs) have been published, which mainly include RIP-seq, RIP-ChIP-seq, and RIP followed by RT-qPCR.To study the effect of ADARs on RNA editing, RIP-seq for high-throughput analysis for target RNA identification can be performed. 165A similar screening approach, using high-throughput RNA-seq after ADAR1-RIP, showed the interaction of ADAR1 with highly edited lncRNAs. 10Furthermore, to study a specific RNA transcript as target, RIP followed by RT-qPCR could be used.RIP-RT-qPCR involves immunoprecipitation of the RBP and performing targeted RT-qPCR on the co-precipitated RNA. 165For instance, in HEK293 cells, ADAR2-specific RIP confirmed its interaction with a pre-mRNA AKAP13, leading to reduced AKAP13 RNA stability and circularization in failing human hearts. 163 further study the downstream effects of the deaminases, knockin and knockdown experiments using plasmids, siRNAs, viral vectors, or CRISPR-based techniques can be used to mimic effects of ADAR or APOBEC modulation on RNA editing events.For example, an siRNA against ADAR1 or ADAR2 in human umbilical cord fibroblast cells was used to determine the effects of ADAR1 and/or ADAR2 expression on editing of both primary and mature microRNAs under ischemic conditions. 26Even though siRNAs have been shown to effectively silence ADARs, the off-target effects of silencing such as modulation of editing patterns in other genes or changes in gene expression via miRNA processing should be noted.In this regard, other approaches might be useful.ADAR2, for example, besides its active site, binds to its target RNA via several residues present in three main regions close to the RNA editing sites. 166As such mutating the specific residues could help in targeted mimicking of RNA editing, thereby reducing the off-target effects.A similar approach can be applied to modulate APOBEC expression to study editing of both coding and ncRNAs. 156,167Thus, studying the role of deaminases in the regulation of edited coding or ncRNAs can provide a foundation for further functional studies.

Experimental induction of RNA editing
For functional analysis, it is crucial to develop methods for generating the unedited and edited versions of the target RNA Figure 6D).Recently, efficient CRISPR-based tools have been developed to achieve this, which involve recruitment of the endogenous editing enzymes to specific sites in RNAs, using guide RNAs, to create edited transcripts.An example for A-to-I editing is RESTORE (recruiting endogenous ADAR to specific transcripts for oligonucleotide-mediated RNA editing). 168RESTORE involves engineering of anti-sense oligonucleotides (ASOs) containing chemically active ADAR recruiting domains.Upon transfection into different human cell types, these ASOs could recruit endogenous ADAR, causing editing of endogenous transcripts while maintaining the natural editing homeostasis. 168Another system, using ADAR1/2 enzymes along with ADAR specific guide RNAs, has been developed for targeted A-to-I RNA editing in both in vitro and in vivo models. 169other RNA editing method called LEAPER (leveraging endogenous ADAR for programmable editing of RNA) uses engineered ADAR recruiting RNAs (arRNAs) to recruit native ADAR1 or ADAR2 to facilitate A-to-I transition on target RNAs.Recently, an updated version of LEAPER, called LEAPER 2.0 has been described.LEAPER 2.0 uses circular ADAR recruiting RNAs (circ-arRNAs) to achieve precise RNA editing with higher efficiency and less off-target effects in both in vitro and in vivo conditions. 170in et al. generated both fully unedited and fully edited Filamin A mice, leading to a Q-to-R amino acid change.These mice were then used to functionally validate the role of Filamin A editing in both tumor and ischemia models. 171For C-to-U editing, Bhakta et al. engineered an RNA editing enzyme complex by using an MS2-tagged system and a combination of the deaminase domain of APOBEC1 with a guide RNA, complementary to the blue fluorescence protein mRNA, creating a C-to-U point mutation, which resulted in a green fluorescence protein. 159ese model systems with unedited and edited RNAs can be used to investigate the role of RNA editing in disease progression in multiple disease phenotypes.For example, using 3D spheroid assays, proliferation assays, transwell cell migration assays, and hindlimb ischemia assays, Jain et al. demonstrated the role of Filamin A mRNA editing in angiogenesis, tumor growth, metastasis, and post-ischemic blood flow recovery, respectively, in a murine hindlimb ischemia model. 171cusing on C-to-U editing, knockout of APOBEC1 was shown to be associated to severity of experimental autoimmune encephalomyelitis, a mouse model mimicking multiple sclerosis.Using techniques such as hematoxylin and eosin staining and immunostaining, Dafou et al. demonstrated the role of APOBEC1 in causing increased inflammation, gliosis, astrocytosis, and T cell infiltration in the brain of APOBEC1 knockout mice presenting the role of C-to-U RNA editing in neurological diseases. 167As such, functional assays appropriately designed for the disease type can provide additional evidence on the impact of editing of diverse RNA species on disease phenotype.
A more direct approach is feasible when studying small RNAs.In two different studies, van der Kwast et al. looked at vasoactive microRNAs that are edited under ischemia.MicroRNAs of the 14q32 gene locus play a directing role in vascular remodeling and in the adaptive vascular response to ischemia.In their studies, van der Kwast et al. show that ischemia induces editing of several 14q32 microRNAs, which alters their targetome and leads to an enhanced pro-angiogenic activity.Using commercially available microRNA mimics, the authors induced overexpressing of either the unedited or edited micro-RNAs.Using different functional assays, the authors demonstrated that the unedited microRNAs inhibit angiogenesis, whereas the edited microRNAs induce angiogenesis, both in vitro and ex vivo. 26,27inical relevance Dysregulation of RNA editing has been implicated in several diseases including cardiovascular and neurological diseases. 11,172,173An effective way to establish clinical impact is to study global editing events and their influence on gene expression in patient samples (Figure 6E).
In the cardiovascular disease context, using ex vivo cultured human arteries and veins, van der Kwast et al. showed an association of mi-croRNA editing with peripheral artery disease, by demonstrating active regulation of microRNA editing during ischemia. 26In a study that aimed to delineate the role of ADARs in congenital heart disease, Altaf et al. found a downregulation of ADAR2 in 35 whole-blood samples and left ventricular tissues of patients with dilated cardiomyopathy compared with 13 healthy controls, suggesting a role of ADARs in cardiovascular disease pathogenesis. 174Considering these findings, RNA editing enzymes such as ADARs may be promising biomarkers in cardiovascular diseases, as is further discussed below.
In the neurological/neurodegenerative disease context, decreased levels of global A-to-I editing were observed in postmortem brain tissues of 55 patients with Alzheimer's disease, compared with 44 nondemented controls. 172Besides Alzheimer's, editing is associated with Creutzfeldt-Jakob disease and other neurodegenerative disorders. 13,175,176Furthermore, multiple studies have shown decreased ADAR2-mediated RNA editing in the motor neurons of patients with amyotrophic lateral sclerosis (ALS).In this regard, restoring ADAR2 activity and targeting dysregulated RNA editing may be a promising novel approach for future ALS therapy. 177sides developing therapeutics directed at editing enzymes or at specific editing events, it has been shown that RNA editing can influence sensitivity to certain drugs, such as a group of synthetic compounds known to act as open-channel blockers that target voltage-activated potassium (Kv) channels in neurons. 178Via such mechanisms, RNA editing may offer a challenge in (future) drug development but may also offer opportunities toward personalized medicine.As such, along with the development of RNA editing drugs, it is imperative to define their safety and effectiveness in the specific cell types where they are intended to be used.Moreover, innovative approaches to guide the drugs to the cell type of interest and methods to follow the effects of these drugs are needed.

Biomarkers
As briefly mentioned above, ADAR2 expression, but potentially also editing events themselves, may have potential as disease biomarkers.To determine the validity of a biomarker, a commonly used statistical test is the area under the curve (AUC) of a receiver operating characteristic plot.An AUC of >0.8 indicates a good diagnostic discrimination. 179In a recent study, Salvetat et al. showed that whole-blood RNA editing variant detection as potential biomarker for diagnosing patients with depression (n = 267) compared with controls (n = 143) has an AUC of 0.930.In addition, a combination of six RNA editing related biomarkers were able to differentially diagnose patients with unipolar (n = 160) vs. bipolar disorder (n = 95; AUC = 0.935). 180pact and translational/therapeutic perspectives RNA editing has been revealed as an additional level of complexity in gene regulation and dysregulation.This regulation affects both short and long RNAs, protein-coding RNAs, and ncRNAs.Changes in nucleotide sequences have been shown to be regulated in several pathological conditions, hence constituting an interesting and relatively novel field of investigation.Discovery of associations between changes in RNA editing and disease development and progression has the capacity to bring not only an enhanced knowledge of the molecular mechanisms beyond these associations, but also to unravel novel therapeutic targets and biomarkers.Indeed, with the rapid evolution of RNA-seq techniques, the possibility to "repair" genes or RNA transcripts using CRISPR-based technology, or the feasibility to measure circulating levels of edited RNA transcripts, RNA editing studies may well lead to the testing of novel treatments and diagnostic strategies.These strategies can be applied to any disease, not only brain and heart diseases, which have been taken as examples in the present article.In favor of RNA editing as a therapeutic tool, in vivo preclinical studies using ADAR-mediated RNA editing to correct missense and nonsense mutations have been reported.In two Rett syndrome mouse models, ADAR was used to repair the Mecp2 mutation and protein function. 181In another mouse study, cardiac-specific ADAR2 overexpression protected against doxorubicin-induced cardiotoxicity and decreased cardiac injury and fibrosis upon acute myocardial infarction, suggesting that ADAR editing may be a promising therapeutic strategy for heart diseases. 182Mechanistically, ADAR2 stimulated neonatal cardiomyocyte proliferation and inhibited doxorubicin-induced cardiotoxicity by affecting the maturation of miR-34a primary transcript and regulating the expression of its target genes Sirt1, Cyclin D1, and Bcl2.
Several Biotech and Pharma companies aim to leverage RNA editing for novel therapeutic strategies, in the brain and heart fields, but not only in these fields.It is expected that these companies will allow translating research findings to clinical application.However, many challenges remain to be solved before applying RNA editing as novel therapeutics, such as specificity, dosage, safety, and so on.Besides the "common" challenges that affect all novel therapeutic strategies, an additional, and perhaps hard to overcome, challenge for therapies that target RNA editing may be the extreme spatiotemporal specificity of both RNA expression and RNA editing.When looking at micro-RNAs, for example, we know that their expression levels can vary greatly within the same organ or tissue, 183 that their target mRNAs are different from cell type to cell type within a single organ or tissue, 184 and also that editing levels of the exact same editing site can vary greatly between different cell types within the same tissue. 26,27his implies that an extremely detailed understanding of spatiotemporal RNA editing is required for any editing-based therapeutic strategy to be both safe and effective, underlining the importance of the complex methodology discussed above.
As for any novel field of investigation, the research on RNA editing has a strong potential to generate intellectual property rights, innovative drugs, and molecular diagnostic tests.Care must be taken when designing research projects, to make sure that the developed tools are clinically translatable.Involving end-users in research projects, from their inception (grant application level) to their development and finalization, is crucial to ensure adoption of developed tools and translation to the clinic.Not only clinicians and private commer-cialization partners, but also patient organizations should be involved in every step of a project, from its very early inception/design phase, to its implementation or testing phase by end-users, which are patients overall (https://patient-engagement.eu/).Engaging patient organizations as full partners in research projects ensures a higher level of translatability, and adoption of novel solutions, for the individual patient's benefit.Especially with mechanisms such as RNA editing, that has been linked to very tight, tissue-specific regulation and to individual drug (in)sensitivity, the potential for both biomarkers and personalized medicine seems very high.

CONCLUSIONS
Even though there are still considerable challenges in RNA editing research, we present a complete study and analysis pipeline here that enables RNA editing studies in biological samples all the way from RNA editing site discovery, through validation and functional analysis, to clinical relevance.The examples used here, taken from the cardiovascular and neurological/neurodegenerative disease fields, underline the importance of RNA editing research.Novel tools are continuously being developed that will facilitate future RNA editing studies even more.
It should be noted that RNA editing comprises just 2 of over 170 different types of RNA modifications that have been discovered to date.Therefore, we look forward to a great number of highly exciting RNA modification studies in the near future.

Figure 1 .
Figure 1.Experimental pipeline for RNA editing studies (A) Summary of the process of sample collection, storage, and RNA isolation, depending on the type of sample.(B) The required steps for RNA-seq library preparation, and the existing RNA-seq approaches depending on the average size and type of RNA.(C) Summary of the main existing bioinformatic tools for RNA editing analysis.(D) Summary of the workflow of utilizing software tools for the identification of potential biological function.(E) Summary of the workflow for experimental validation of RNA editing events, using different PCR techniques.(F) Summary of various experimental methods for the detection of functional effects of RNA editing events, and confirmation of clinical relevance of RNA editing in patient samples.Created with BioRender.com.

Figure 2 .
Figure 2. Steps for plasma handling and processing Several crucial steps and pitfalls are highlighted to both warrant high-quality RNA yields from plasma and circulating extracellular vesicles and to avoid hemolysis and platelet contamination.Created with BioRender.com.

Review
need to utilize SNP databases, by clustering RES-based and SNPbased single-nucleotide variant (SNV) duplets, due to their distinctive and unique distribution.In more detail, A-to-I and C-to-U editing are inescapably being observed in different genomic regions, since ADAR enzymes (responsible for A-to-I editing) act only on double-stranded RNA, while APOBEC enzymes (responsible for C-to-U editing) act only on single-stranded RNA.In addition, it is known that ADARs tend to edit sites in clusters (hyper RNA editing) and are prevalent in intronic regions and Alu repeats,20,21,[108][109][110][111][112] whereas SNPs present independent distribution of each SNP type and low density in the genome.SPRINT is fully automated to any RNA-seq data with available reference genome and can use both raw FASTQ files or BAM files as input.It can detect both de novo and known editing sites and can also integrate the detection of hyper RNA editing sites from remapped reads.

Figure 4 .
Figure 4. Schematic illustration for JACUSA2 workflow analysis JACUSA2 tool predicts single-nucleotide differences from both RNA-DNA (RDDs) and RNA-RNA (RRDs) sequencing data.PCR duplicates are being marked and variants in the start and end of the genome, in INDELs, in intronic regions and in homopolymers are being filtered accordingly to prevent false-positive RNA editing events.Created with BioRender.com.

Figure 5 .
Figure 5. Schematic illustration for SPRINT bioinformatics analysis SPRINT enables the detection of A-to-I and C-to-U editing sites by mapping reads to the corresponding reference genome using Burrows-Wheeler aligner and clustering RES-based and SNP-based SNV duplets.Both regular RES from mapped reads and hyper RES from recovered remapped reads can be identified.Created with BioRender.com.

Table 1 .
Main features of bioinformatic tools for RNA editing detection Molecular Therapy: Nucleic Acids Vol.35 March 2024 www.moleculartherapy.org