Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Apr 22.
Published in final edited form as: Methods Enzymol. 2025 Jan 2;710:131–152. doi: 10.1016/bs.mie.2024.11.030

En masse evaluation of RNA guides (EMERGe) for ADARs

Prince J Salvador 1, Natalie M Dugan 1, Randall Ouye 1, Peter A Beal 1,*
PMCID: PMC12014283  NIHMSID: NIHMS2071118  PMID: 39870442

Abstract

Adenosine Deaminases Acting on RNA (ADARs) convert adenosine to inosine in duplex RNA, and through the delivery of guide RNAs, can be directed to edit specific adenosine sites. As ADARs are endogenously expressed in humans, their editing capacities hold therapeutic potential and allow us to target disease-relevant sequences in RNA through the rationale design of guide RNAs. However, current design principles are not suitable for difficult-to-edit target sites, posing challenges to unlocking the full therapeutic potential of this approach. This chapter discusses how we circumvent this barrier through an in vitro screening method, En Masse Evaluation of RNA Guides (EMERGe), which enables comprehensive screening of ADAR substrate libraries and facilitates the identification of editing-enabling guide strands for specific adenosines. From library generation and screening to next generation sequencing (NGS) data analysis to verification experiments, we describe how a sequence of interest can be identified through this high-throughput screening method. Furthermore, we discuss downstream applications of selected guide sequences, challenges in maximizing library coverage, and potential to couple the screen with machine learning or deep learning models.

1. Introduction

1.1. ADAR and RNA editing

ADARs (Adenosine Deaminase acting on RNA) are enzymes that catalyze the hydrolytic deamination of adenosine (A) to inosine (I) in dsRNA (Wulff & Nishikura, 2010). Inosine is recognized by many downstream cellular processes as guanosine, thus an A-to-I conversion can result in a codon change (Bass, 2002). While ADARs require double stranded substrates, it is not necessary for the RNA to be perfectly Watson-Crick base paired (Schneider et al., 2014). In fact, select mismatches have been shown to support higher levels of editing (Bass, 2002; Quiroz et al., 2023). Bulges and loops also have a demonstrated effect on selectivity (Hundley & Bass, 2010; Lehmann & Bass, 1999). A mismatching cytidine directly across from the edited adenosine has been shown to improve editing efficiency (Wong et al., 2001). In addition, the immediate flanking sequence context around an editing site has been shown to be impactful. The 3’ and 5’ bases flanking the target adenosine are well characterized determinants of substrate preferences (Eggington et al., 2011; Matthews et al., 2016). Also, select mismatches, such as a G-G mismatch adjacent to an editing site, have been shown to overcome natural preferences to enable editing of non-preferred sequences (Doherty et al., 2022; Schneider et al., 2014).

1.2. Site directed RNA editing (SDRE)

Since ADARs require a duplex RNA structure for binding, antisense guide oligonucleotides can be introduced to specifically base pair with therapeutically relevant transcripts, thereby creating a duplex structure that can recruit endogenous ADAR to select editing sites. This site directed RNA editing (SDRE) approach is rapidly gaining widespread recognition as a therapeutic approach for genetic disorders (Khosravi & Jantsch, 2021; Monian et al., 2022). In a recent example, a direct reversion targeting a pernicious G-to-A mutation in the SERPINA1 gene has been shown to restore up to 75 % of deficient α1-Antitrypsin in mouse models (Monian et al., 2022); an approach that Wave Life Sciences has taken into clinical trials as of April 2024 (Suter, 2024). In addition, work is in progress to repair premature termination codons arising from mutations that cause Rett syndrome (Brinkman et al., 2022; Doherty et al., 2022; Jacobsen et al., 2023; Matos et al., 2024; Pfeiffer & Stafforst, 2023; Sinnamon et al., 2017, 2020, 2022; Wong et al., 2024). Beyond addressing disease related nonsense mutations, ADAR-based mRNA editing has the potential to yield amino acid recoding events altering cell physiology, functions and stability of enzymes or their interactions with substrate (Quiroz et al., 2023).

1.3. Screens for guide strand sequences that enable editing at recalcitrant sites

Antisense guide oligonucleotide design is paramount in the effort to redirect endogenous ADAR to a therapeutically relevant mRNA sequence, particularly in those sequence contexts which ADAR does not naturally prefer. Notably, sequences with a guanosine 5’ of the target adenosine are known to edit poorly, as described in Section 1.1. The ideal guide oligo promotes efficient ADAR binding and editing, but the design principles and structural nuances that dictate ideal guide sequence are incompletely understood. Employing a high-throughput guide sequence screening strategy, identifying sequence motifs supporting high editing efficiency, can emerge that push forward the boundaries of current understanding. For example, a guide sequence motif discovered by the methodology described herein has been shown to double in vitro deamination levels relative to the prior reported designs (Jacobsen et al., 2023). In another recent exciting example from the Rosenthal lab, a screening methodology similar to the one discussed here indicates that while complementarity is preferred in some duplex positions (−5, −4, −1, +1, +2 and +3), relative to edited site 0, other sites slightly further from the editing site (−8, −6, +10, +11) demonstrate a preference for non-complementarity (Quiroz et al., 2023). Meanwhile, a SELEX based screening method produced evidence that other non-Watson-Crick base pairing perturbations that are tolerated by ADAR can be introduced to suppress off-target edits (Wong et al., 2024). Herein we provide a detailed protocol for the EMERGe (En Masse Evaluation of RNA Guides) screening strategy, in which a section of a guide oligonucleotide is sequence randomized and the resulting library is screened to determine which sequences support high levels of editing (Jacobsen et al., 2023) (Fig. 1).

Fig. 1.

Fig. 1

Open in a new tab

An overview of the EMERGe workflow. The ssDNA starting material is synthesized chemically via the phosphoramidite method to include T7 promoter, target adenosine, hairpin linker, and randomized N10 library. After amplification via PCR the now dsDNA is transcribed with T7 RNA polymerase to form the EMERGe RNA hairpin. The hairpin is subjected to deamination by ADAR then reverse transcribed to enable NGS readout of N10 sequences and their associated levels of adenosine deamination. These data are processed and tabulated to discern library sequences that support high levels of deamination. Created with Biorender.com.

The EMERGe library screening method hinges on a hairpin structure in which one arm models the disease relevant target sequence in the associated RNA transcript. An opposing antisense arm models the guide oligo, with a specific number of nucleotides across from the editing site randomized to constitute the sequence library. In our initial EMERGe library design, ten nucleotides opposite the target editing site were randomized (N10 sequence). Importantly, the two arms are covalently linked by an oligonucleotide loop to form a hairpin duplex substrate. In this way, the hairpin library may be deaminated by ADAR in vitro, reverse transcribed to DNA and analyzed by Next Generation Sequencing (NGS). Because the target adenosine and N10 sequence are present on the same DNA molecule after reverse transcription, the sequence data can be queried such that adenosine editing percentage can be correlated with each sequence in the 10-nucleotide randomized region. To verify that winning guide motifs continue to support high levels of editing outside the hairpin context of the model system, new guide oligos are synthesized containing said motifs and tested in trans paired with a longer (~300nt) single stranded RNA target. The resultant duplex is subjected to in vitro deamination with ADAR and editing efficiency is determined (Jacobsen et al., 2023). Once confirmed as an editing-enabling motif, the novel structure can be inserted into ADAR guide strands for various SDRE applications.

2. EMERGe screening protocol

2.1. Equipment and programs

Access RT-PCR Kit (Promega).

Agarose gel box and power supply.

AKTA® Pure, or equivalent FLPC and associate fraction collector.

Benchtop Centrifuge.

Bunsen Burner.

Centrifuge with 6 L rotor.

Digital heat block, 30 °C.

Digital heat block, 95 °C.

Freezer, −70 °C.

Incubator shaker, CO2 control, 37 °C.

Microfluidizer.

Micropipettes.

Ni-NTA column (Thermofisher).

Nutating rocker.

Pasteur pipettes and electronic pipettor.

PCR Thermocycler (BioRad, or equivalent).

Plate Reader or UV spectrophotometer.

Plate spreader and loop.

PAGE Gel stand and power supply.

Refrigerator, 2–4 °C.

SDS-PAGE gel box and power supply.

SpeedVac (Thermo), or equivalent.

2.2. Materials

0.2 μm centrifugal filters.

1 M Tris-HCl, pH 7.4.

1 M Tris-HCl, pH 8.0.

100 bp Agarose Ladder (New England Biosciences).

6x Agarose loading dye (New England Biosciences).

10x TBE buffer (Corning).

15 mL and 50 mL falcon tubes.

2 L Bellco Flasks.

2 mL microcentrifuge tubes.

Access RT-PCR Kit (Promega).

Agar.

Agarose (Qiagen).

Acrylamide.

β-mercaptoethanol.

bis-Acrylamide.

Bovine Serum Albumin (Thermo Scientific).

Custom Gene Fragment (Integrated DNA Technologies).

Dextrose.

Dithiothreitol.

DNase I and DNase Buffer (Thermofisher).

Ethylenediaminetetraacetic acid.

Ethanol, 200 proof.

Galactose.

Glycerol.

HiScribe® T7 High Yield RNA Synthesis Kit (New England Biolabs).

Imidazole.

Lactic acid.

Magnesium Chloride.

Minimal Media -Uracil.

Nonidet P-40 (Thermofisher).

Sodium Chloride.

Petri Dishes.

Phusion® High-Fidelity Polymerase Kit (New England Biolabs).

Potassium Chloride.

Purified hADAR2 (see above).

Qiaquick gel extraction kit (Qaigen), or equivalent.

RNA Library Hairpin (see above).

RNase Inhibitor (Invitrogen).

Sodium acetate.

S cerevisiae (BCY123).

SYPRO Orange (Invitrogen).

Tris-HCl, pH 7.8.

TritonX-100 (Sigma Aldrich).

Urea.

Yeast tRNA (Invitrogen).

2.3. Hairpin design and library preparation

  1. Libraries are synthesized via phosphoramidite chemistry or purchased via Integrated DNA Technologies (IDT) as ssDNA with the addition of a 5’ T7 RNA polymerase promoter sequence (Fig. 1).

  2. PCR amplification (NEB M0530L) produces dsDNA, which is then purified on a 1 % agarose gel. PCR amplification is performed in a total of 25 cycles: (1) denaturing at 95 °C for 30 s, (2) annealing at 60 °C for 1 min and (3) extension 68 °C for 2 min

  3. Purified amplicon is subjected to T7 RNA polymerase in vitro transcription (NEB E2040) to produce the ssRNA library according to kit instructions. The ssRNA library is DNase I treated (Thermofisher, AM2222) and purified by denaturing 12 % PAGE. Target bands are excised, crushed and eluted by soaking overnight in 0.5 M NaOAc, 0.1 mM EDTA at 2–4 °C.

  4. Crushed PAGE gel fragments are removed from sample by 0.2 μm filter and RNA is precipitated from a supernatant by dilution with EtOH to 75 % (v/v) −70 °C for 12 h and 17,000 × g centrifugation. Precipitate is lyophilized to dryness.

  5. Dry sample is reconstituted in 100 mM NaCl, 1 mM EDTA, and 10 mM Tris pH 7.4. The RNA is folded by denaturation at 95 °C and slowly cooled to RT over 2 h.

2.4. ADAR overexpression and purification

  1. 10 xH is tagged hADAR2 protein is overexpressed in 2L S. cerevisiae culture as described (Macbeth & Bass, 2007).

  2. Yeast cells harboring overexpressed hADAR2 are pelleted and lysed via microfluidizer in 50–100mL lysis buffer (20 mM Tris-HCl pH 8.0, 5 % (v/v) glycerol, 750 mM NaCl, 35 mM imidazole, 0.01 % (v/v) Triton X-100 and 1 mM β-mercaptoethanol (BME)). The lysate is centrifuged (39 000 g for 1 h).

  3. The supernatant is loaded onto 5 mL Ni-NTA column, washed once with 50 mL lysis buffer, then with 50 mL each of the following wash buffers: (1) 20 mM Tris-HCl pH 8.0, 5 % (v/v) glycerol, 300 mM NaCl, 35 mM imidazole, 0.01 % (v/v) Triton X-100 and 1 mM BME; then (2) 20 mM Tris-HCl pH 8.0, 5 % (v/v) glycerol, 100 mM NaCl, 35 mM imidazole, 0.01 % (v/v) Triton X-100 and 1 mM BME. The column is then eluted with a gradient of 30 to 400 mM imidazole in 20 mM Tris-HCl, pH 8.0, 5 % glycerol, 1 mM BME, 100 mM NaCl.

  4. After analysis by SDS-PAGE, combined fractions are dialyzed into 20 mM Tris-HCl pH 8.0, 20 % (v/v) glycerol, 100 mM NaCl and 1 mM BME.

  5. Protein concentration is determined relative to bovine serum albumin (BSA) standards on an SDS-PAGE gel, followed by SYPRO Orange (Invitrogen) staining.

2.5. In vitro ADAR reaction with RNA library

  1. Human ADAR2 (200 nM) is mixed with 10 nM RNA hairpin in 15 mM Tris-HCl, pH 7.8, 60 mM KCl, 3 mM MgCl2 1.5 mM EDTA, 3 % glycerol, 0.003 % Nonidet P-40, 0.6 mM DTT, 160 U/mL RNase inhibitor, and 1.0 μg/mL yeast tRNA in a final volume of 100 μL and allowed to react for 30 min at 30 °C. The reaction is then quenched pipetting into 900 μL of 95 °C water and held at this temperature for 5 min, then chilled on ice for 3 min

2.6. Illumina sequencing

  1. The post-deamination RNA is concentrated to 350 μL in vacuo and subjected to RT-PCR (Promega A1280).

  2. The AMV reverse transcriptase step is performed for 45 min at 45 °C and the resultant cDNA is amplified by Tfl polymerase for 24 cycles (1 min at 60 °C annealing, 2 min at 68 °C extension, 30 s at 94 °C melting).

  3. The amplified cDNA was purified by 1 % agarose gel (Qiagen 28706) and excised bands are extracted via silica column (Qiagen 28704) and dried in vacuo.

  4. The pellet is resuspended in nuclease free water and 2.5 μg of sample is submitted for library preparation and Illumina next-generation sequencing. Samples (20 μL, 50 ng/μL) are submitted to Azenta for the following service type: premade amplicon sequencing without fragmentation, single index, 150 bp 350 M PE reads with 30 % Phi-X Spike.

3. Analysis of EMERGe screening results

The script for MiSeq data pre-processing and data processing have been published from our laboratory (https://github.com/csjacobsen/EMERGe/tree/bioinformatics) (Jacobsen et al., 2023).

3.1. Equipment

Computer.

Linux.

R-Studio.

Microsoft excel.

3.2. Bioinformatic analysis of sequencing data from screens

  1. Linux software is used for pre-processing raw sequencing files with the code provided in (https://github.com/csjacobsen/EMERGe/tree/bioinformatics). This step utilizes a HTSStream application, a toolset that acts as a processing pipeline to prepare data analysis on the R studio application. It is imperative that the Stats, SeqScreener, Overlapper, Primers, NTrimmer, and LengthFilter commands are executed to ensure appropriate data will be taken forward for analysis (https://s4hts.github.io/HTStream/).

  2. The pre-processed fastq file is exported to designated processing folders in preparation for further data processing.

  3. R-studio is used for the processing code provided in (https://github.com/csjacobsen/EMERGe/tree/bioinformatics) as a template. The library sequence from 5’- > 3’ is inserted in the “DNA String” line, and followed by the desired sequence length.

  4. The subseq(sread(cjr)) corresponds to the five nucleotides flanking the 5’ end of the target (21–26), five nucleotides flanking the 3’ end of the target (30–35), 5’ end of the N10 library (60–65), and 3’ end of the N10 library (76–81) (Fig. 2). Locations of the flanking sequences and library can vary depending on hairpin design.

  5. After executing the script, a csv file output will be available for further processing.

  6. In the pre-processed csv file, all three nucleotide “codons” that include the target A except the unedited codon (NAN), and edited codon (NGN) are deleted, where the underlined target adenosine and N is A, C, G or U. Combining the reads for the unedited and edited codon results in the calculated read count for the guide sequence.

  7. The editing percentage is calculated using the following equation: (reads of edited codon/reads of edited codon + reads of unedited codon) × 100 %.

  8. Using the filter function, implement a read number and editing level cut off, and rank the N10 sequences from highest to lowest percent editing.

Fig. 2.

Fig. 2

Open in a new tab

(A) Example 96 nucleotide EMERGe library design. Highlighted in green are regions used in data processing script and highlighted in purple is the library region. (B) Mock results after running processing code and further data processing. The top N10 sequences with the respective GAG and GGG reads of the target A is ranked based on percent editing.

3.3. Single nucleotide variant analysis of top scoring hits

The initial analysis of the NGS data from an EMERGe screen can identify guide sequences associated with high levels of A to G conversion at the target site. An additional layer of analysis can be applied for high performing guide sequences, where the dataset is queried for single nucleotide variants. This in silico analysis identifies which nucleotides can be varied and which nucleotides are highly sensitive to a nucleotide change in the library region, therefore illuminating sequence motifs essential for enabling efficient editing at the predetermined target site.

3.3.1. Procedure

On Microsoft excel, all filters are removed in processed files, a filter function to the guide sequence column is added, and each position of the N10 of a specific guide sequence is queried by searching (from 5’–3’): “?NNNNNNNNN”, where N is the identified guide sequence and ? is the query code input. Single nucleotide variants found in the dataset will populate alongside their respective percent editing.

An activity heat map is generated from single nucleotide variants of a specific sequence of interest (Fig. 3).

Fig. 3.

Fig. 3

Open in a new tab

Example of a heat map generated from a single nucleotide variant analysis of a guide RNA. 5’ to 3’ target sequence with the target A in red and 3’ to 5’ candidate guide shown above heat map. Created with Biorender.com.

3.3.2. Validation of hit sequences in ADAR editing assays

The covalent linkage of the target sequence with the gRNA library through a hairpin loop in an EMERGe screen is not fully representative of an SDRE application, where a guide sequence is hybridized with a target strand, and thus serving as a substrate for an ADAR reaction. The presence of the hairpin opens the possibility for the screen to identify guide sequences that are highly reactive to ADAR, but are dependent on the hairpin loop. To validate the hit sequences from an EMERGe screen that can be useful in SDRE, one must generate oligonucleotide guide strands bearing these sequences that target the site of interest and test their editing efficiency in follow up ADAR assays.

3.4. Equipment

Benchtop centrifuge.

Freezer, −70 °C.

Heat block, 30 °C.

Heat block, 95 °C.

Pipettes.

Plate reader or UV-Vis Spectrophotometer.

Thermal cycler machine (Biorad or similar).

3.5. Materials

100 bp ladder (New England Biosciences).

15 mL and 50 mL falcon tubes.

2 mL microcentrifuge tubes.

Access RT-PCR (Promega).

Acrylamide.

bis-Acrylamide.

Customized gene block encoding target sequence (IDT).

Customized ssDNA encoding guide sequence and HDV ribozyme sequence (IDT).

DNA Clean & Concentrator Kit (Zymo Research).

DNase I and DNase Buffer (Invitrogen).

DTT (dithiothreitol).

Ethylenediaminetetraacetic acid (EDTA).

Glycerol.

HiScribe T7 High Yield RNA Polymerase (New England Biosciences).

6x loading dye (New England Biosciences).

Nonidet P-40 (Thermofisher).

PCR tubes.

Phusion High Fidelity DNA Polymerase (New England Biosciences).

Primer sequences (IDT).

Potassium Chloride.

Purified hADAR enzyme (see 2.2).

RNase Inhibitor (New England Biosciences).

Sodium Chloride.

Tris-borate-EDTA (Corning).

Urea (Thermo Scientific).

Yeast tRNA (Thermo Fisher Scientific).

Our laboratory has published verification experiments with guide sequences synthesized using T7 RNA polymerase as it was cost-effective and utilizes a well-established methodology (Jacobsen et al., 2023). However, as the number of candidate EMERGe hit sequences to be tested in our laboratory increased, we were confounded by the purification problem arising from the ability of T7 RNA polymerase to catalyze 3’ end additions, leading to mixtures of products of similar length (Gholamalipour et al., 2018). To address this, we chose to incorporate a ribozyme sequence that catalyzes cleavage at the 3’ end of the transcript to yield a homogeneous product of defined length. Indeed, we were able to achieve this by inducing a co-transcriptional 3’ end cleavage using HDV56 (Milligan et al. 1987) (Fig. 4). This procedure leads to a single major guide RNA product whose mass is readily confirmed by MALDI-MS analysis. Significantly, this method for generating candidate guide RNAs for validation experiments is cost effective (20 guides for ~ $500) and substantially more economical compared to an estimated $6000 when purchasing 20 chemically synthesized 30 mer guide RNAs from a commercial vendor.

Fig. 4.

Fig. 4

Open in a new tab

In vitro co-transcriptional 3’ cleavage with HDV sequence and T7 RNA Polymerase for gRNA synthesis. Created with Biorender.com.

3.6. Generation of 30 nt candidate guide strands bearing selected sequences using in vitro transcription

3.6.1. Procedure

  1. Two ssDNAs are hybridized: (A) selected guide sequences can be generated through a DNA strand encoding (from 5’- > 3’), a T7 promoter sequence and the 30 nt guide sequences containing the N10 library identified from the screen and (B) reverse complement HDV sequence upstream of the complementary region 3’ end of the ssDNA A. The dsDNA is hybridized through heating at 95 °C for 5 min and cooling to room temperature under the following conditions: 5 μM DNA, 1 mM EDTA, 10 mM Tris pH 7.4.

  2. The hybrid is amplified through PCR following the manufacturer’s protocol to produce a dsDNA containing, from 5’- > 3’, a T7 promoter sequence, guide sequence containing N10, and HDV ribozyme sequence, respectively. The product is column purified using a DNA Clean & Concentrator Kit following the manufacturer’s protocol.

  3. dsDNA (2 μg) is used as a transcription template for T7 RNA polymerase following the manufacturer’s protocol, and purified on a denaturing PAGE gel. We have found that 2 μg of transcription template per 20 μL reaction is sufficient to yield ample guide RNA product post 3’ cleavage.
    Component Amount
    NTP Buffer Mix (10 mM each NTP) 10 μL
    Template DNA 2 μg
    T7 RNA Polymerase Mix 2 μL
    Open in a new tab

  4. Gel bands are excised, crushed and soaked in 0.5M NaOAc, 0.1 mM EDTA at 4 °C overnight. Gel fragments are removed by 0.2 μm filter and RNA is precipitated from a supernatant to a final EtOH concentration to 75 % (v/v) −70 °C for 24 h and 13,000 × g centrifugation.

  5. The identity of each guide strand to be tested is confirmed using mass spectrometry.

3.6.2. Table of sequences

Sequence
T7 RNA Polymerase Promoter Sequence 5’-TAATACGACTCACTATAGGG-3’
HDV Ribozyme Sequence 5’-GAGGGATAGTACAGAGCCTCCCCGT
GGCTCCCTTGGATAACCAACTGATACT
GTAC-3’
Open in a new tab

3.7. ADAR deamination assays with candidate guides

A ~ 300 nt gene fragment with an inserted T7 promoter sequence upstream the sequence of interest is amplified through PCR following the manufacturer’s protocol. The product is purified using a DNA Clean & Concentrator Kit as described in 4.3.1.

The purified gene fragment (1 μg) will serve as a substrate for an in vitro T7 RNA polymerase transcription following the manufacturer’s protocol, and purified on a denaturing 18 % PAGE gel.

Gel bands are excised, crushed and soaked in 0.5 M NaOAc, 0.1 mM EDTA at 4 °C overnight. Gel fragments are removed by 0.2 μm filter and RNA is precipitated from a supernatant to a final EtOH concentration to 75 % (v/v) −70 °C for 24 h and 13,000 × g centrifugation.

Each guide strand is hybridized to the target strand to a final volume of 50 μL with a nuclease free water by heating to 95 °C for 5 min and cooled to room temperature for 2 h under the following conditions: 100 nM target strand, 1 μM guide strand, 500 mM NaCl, 1 mM EDTA and 10 nM TrisHCl pH 7.4.

The protocol for the ADAR activity assay for each respective guide is described as: 10 nM hybridized RNA in 15 mM Tris-HCl, pH 7.8, 60 mM KCl, 3 mM MgCl2 1.5 mM EDTA, 3 % glycerol, 0.003 % Nonidet P-40, 0.6 mM DTT, 160U/mL RNase inhibitor, and 1.0 μg/mL yeast tRNA in a final volume of 100 μL. Prior to adding hADAR2 to the reaction, the reaction mixture is incubated at 30 °C for 30 min and as no enzyme control, 10 μL of the reaction is aliquoted into 190 μL of water pre-heated to 95 °C, held at 95 °C for 5 min and cooled on ice. 1 μM of purified ADAR2 is added to the reaction to a achieve a final concentration of 100 nM and 10 μL of the reaction at different time points are quenched into 190 μL of water pre-heated to 95 °C, held at 95 °C for 5 min and cooled on ice.

The manufacturer’s protocol is followed to perform the RT-PCR reaction with target-specific primers.

The RT reaction is performed for 45 min at 45 °C, followed by a 2 min enzyme deactivation step at 94 °C. PCR amplification is performed in a total of 25 cycles: (1) denaturing at 95 °C for 30 s, (2) annealing at 60 °C for 1 min and (3) extension 68 °C for 2 min

The reaction product (2.5 μL) can be analyzed by agarose gel electrophoresis and upon confirmation of desired amplicon, purified using a DNA Clean & Concentrator Kit and submitted for Sanger sequencing analysis.

The editing percentage of the target with the respective guide sequence is calculated using the following equation: [G height/(G height + A height) × 100 % (Fig. 5).

Fig. 5.

Fig. 5

Open in a new tab

Workflow for guide RNA verification experiments. Target and gRNA generated through in vitro transcription and hybridized. Duplex RNA is react with ADAR, product is then subjected through RT-PCR, followed by Sanger sequencing.

4. Application of winning sequences

Once editing enabling sequence motifs have been validated in the experiments described above, they can be used to direct editing with fusion proteins and encodable guide RNAs or in chemically modified editing oligonucleotides (EONs). Furthermore, additional structure-activity relationship analyses and structural studies can advance our understanding of how features present in the RNA substrate contribute to ADAR editing efficiency and selectivity.

4.1. Directed RNA editing with fusion proteins and encodable guide RNAs

The recruitment of ADAR for SDRE has been described by several groups. Fusions of the ADAR deaminase domain with an RNA binding protein and guide RNA, both with Cas13 (Cox et al., 2017; Katrekar et al., 2019) and a λN peptide-boxB (Montiel-Gonzalez et al., 2013, 2016, 2019) serving as the RNA binding protein have been used. In cell lines and primary cells, target RNA substrates have been edited in homogenous cellular muscle and liver tissues (Katrekar et al., 2019), as well as complex central nervous system tissues (Sinnamon et al., 2017, 2020). To verify editing efficiency of EMERGe guides in cellulo, the λN-boxB directed editing system was used to test gRNAs for their ability to correct a premature termination codon in the MECP2 transcript that causes Rett Syndrome (Jacobsen et al., 2023). This study showed that the EMERGe-identified guide strand sequence improved editing yield over a designed guide by approximately two-fold. It is likely that sequences identified in EMERGe screens like those described here will enhance editing yield for encodable ADAR guides in other editing systems such as the recently described CLUSTER or LEAPER systems (Qu et al., 2019; Reautschnig et al., 2022).

4.2. Chemically modified editing oligonucleotides (EONs) containing EMERGe motifs

Therapeutic strategies have also centered on the use of antisense oligonucleotides (AONs) for a variety of disorders and diseases, which delivers an oligonucleotide alone. Editing oligonucleotides (EONs) are chemically modified guide RNAs (gRNAs) that hybridize with a target RNA, creating a double stranded RNA substrate for ADAR to make a therapeutically relevant edit. The RNA is chemically modified for efficient uptake in cells and increased stability of the oligonucleotide (Bost et al., 2021). Also, EONs have less immunogenicity concerns compared to methods that deliver proteins (Stebbins et al., 2019). Guide sequences that have been selected by EMERGe can be chemically modified and delivered as an EON for SDRE in vivo. However, chemical modifications have the potential to impact the ability of EMERGe-identified gRNA to enable editing, so the effect of the chemical modifications must be established independently.

5. Limitations and future work

5.1. Library coverage

Our typical EMERGe screen is carried out with an RNA hairpin bearing an N10 variable region for a library diversity of 1,048,576 individual sequences. We also typically carry out the ADAR reaction using 2.0 μg of RNA hairpin library corresponding to 3.9 × 1013 RNA molecules or > 106 copies of each sequence of an N10 library. The NGS analysis gives a total of 300 million reads corresponding to approximately 300 reads per sequence if all sequences were equally represented. However, with our typical analysis cutoff of at least six reads per sequence, we identify only 600–800K individual sequences in our final data sets. Given the utility of the high-throughput EMERGe screen hinges on the evaluation of a large library of sequences, it is imperative that each step of the screen be considered for where individual sequences may be lost.

DNA library synthesis. The first step that determines library coverage is the generation of the DNA library. The DNA libraries used in Jacobsen et al. (2023) were obtained from Integrated DNA Technologies (IDT), which offers two methods of randomization to create the library: machine- and hand-mix. The machine-mix method allows all four bases to react simultaneously. Due to a difference in phosphoramidite reaction rates, some of the sequences predicted to exist in a truly randomized library may not be present. Hand-mixing can ensure that all sequences in the randomized library are present, at an additional cost.

PCR amplification and transcription. The single stranded DNA library is converted into a duplex by PCR prior to T7 transcription. PCR could bias the distribution of sequences present due to differences in amplification efficiencies, template switching that produces novel chimeras, and polymerase errors (Kebschull & Zador, 2015). This could result in a change in the identity of the DNA templates present or bias sequence populations of the library. The use of T7 RNA polymerase to convert DNA into RNA is another step that could introduce bias. There is a low measured rate for misincorporation by T7 RNA polymerase (Huang et al., 2000), so bias is unlikely to arise from misincorporation. However, it has been shown that the sequence downstream of the T7 promoter can alter the level of RNA output, as GC-rich sequences experience a lower RNA generation compared to AT-rich sequences (Conrad et al., 2020). Since the target sequence that is used for the screen is downstream of the T7 promoter sequence, GC or AT levels of a target sequence has the potential to impact the RNA library output from this step, and thus introduce bias in what sequences are present for the screen.

Reverse transcription. Importantly, once the RNA library has undergone deamination it must be reverse transcribed and PCR amplified prior to sequencing. However, recent reports regarding how RT-PCR can bias the cDNA output (Luas et al., 2023; Verwilt et al., 2023) suggests the RT-PCR step of our EMERGe workflow is a likely source of substantial sequence loss. RNA sequences may not be reverse transcribed based on GC content, transcript length, and/or sequence bias (Minshall & Git, 2020; Oshlack & Wakefield, 2009). The presence of stable secondary structure is a known obstacle for efficient reverse transcription for many RTases. However, the need for ADAR substrates to be double helical and the requirement for the editing site and variable region to be present on the same molecule led us to use a hairpin structure. While we have shown that successful screens can be carried out using AMV reverse transcriptase, the use of thermostable RTases that can tolerate higher temperatures capable of melting stable secondary structures is likely to reduce sequence loss. When comparing thermostability of RTases, one group recommended MMLV-derived SuperScript IV and MaximaH as efficient thermostable RTases (Zucha et al., 2020). In addition, NEB has recently released Induro, a group II intron-encoded RTase exhibiting high processivity and thermostability.

5.2. Defining the mechanistic basis for editing-enabling motifs discovered by EMERGe screens

Validated guide sequences with novel editing-enabling motifs can be chosen for further structure-activity relationship and structural studies. Mismatches present in the EMERGe guide can be replaced with a corresponding Watson-Crick base pair to determine which, if any, mismatches were important for editing efficiency or selectivity. For instance, when an A:A mismatch was replaced with an A:U base pair in an EMERGe-identified guide sequence bystander editing of neighboring adenosines increased indicating a role for the A:A mismatch in controlling editing selectivity (Jacobsen et al., 2023). X-ray crystallography can also be used to define the basis for the effects seen with EMERGe-identified features. Crystal structures provide an atomistic picture of the structural features present in the RNA as well as interactions between protein and RNA. In one of our initial screens targeting a premature termination codon in MECP2 that causes Rett syndrome (R168X), a recurring feature in the highest performing guide strands was a G predicted to pair with a G in the target strand adjacent to the target adenosine (Jacobsen et al., 2023). A G:G pair had previously been shown to enable editing at a 5’ GA site for the SNAP-ADAR editing system as well (Reautschnig et al., 2022). X-ray crystallography was used to show that a specific pairing geometry (Gsyn:Ganti) with the target stand G recognized on its Hoogsteen face by the Watson Crick edge of the guide strand G (Doherty et al., 2022). This insight led to the development of other nucleoside analogs capable of recognizing the target strand G on its Hoogsteen face. Additional structural studies of EMERGe-identified editing-enabling motifs are expected to both advance our understanding of how RNA structure dictates ADAR editing efficiency and lead to novel guide strand features that maximize on-target editing in SDRE applications.

5.3. EMERGe data to generate predictive tools

Each target site for SDRE has a unique sequence context that requires gRNAs to be optimized for efficient and specific editing. By identifying a range of gRNAs for different target sequences, we are building a database of information relating RNA substrate features to ADAR editing efficiency. Exploring the relationship between sequence, structure, and editing will inform guide RNA generation for a wide variety of targets that are not accessible by traditional small molecule therapeutics (Yu et al., 2020). Next generation sequencing (NGS) generates a vast amount of data which can then be used to train machine learning (ML) or deep learning (DL) models (Özgür & Orman, 2023; Schmidt & Hildebrandt, 2021; Zhang et al., 2021) for a variety of applications. The EMERGe screen NGS datasets could be used to train machine learning models to predict ideal gRNA sequences for ADAR editing, especially given data from multiple EMERGe screens could be combined for more complex modeling applications.

References

  1. Bass B (2002). RNA editing by adenosine deaminases that act on RNA. Annual Review of Biochemistry, 71(1), 817–846. 10.1146/annurev.biochem.71.110601.135501. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bost JP, Barriga H, Holme MN, Gallud A, Maugeri M, Gupta D, ... El-Andaloussi S (2021). ACS Nano, 15(9), 13993–14021. 10.1021/acsnano.1c05099. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Brinkman HF, Matos VJ, Mendoza HG, Doherty EE, & Beal PA (2022). Nucleoside analogs in ADAR guide strands targeting 5’-UA sites. RSC Chemical Biology, 4, 74–83. 10.1039/D2CB00165A. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Conrad T, Plumbom I, Alcobendas M, Vidal R, & Sauer S (2020). Maximizing transcription of nucleic acids with efficient T7 promoters. Communications Biology, 3, 439. 10.1038/s42003-020-01167-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Cox DBT, Gootenberg JS, Abudayyeh OO, Franklin B, Kellner MJ, Joung J, & Zhang F (2017). RNA editing with CRISPR-Cas13. Science (New York, N. Y.), 358, 1019–1027. 10.1126/science.aaq0180. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Doherty EE, Karki A, Wilcox XE, Mendoza HG, Manjunath A, Matos VJ, ... Beal PA (2022). ADAR activation by inducing a syn conformation at guanosine adjacent to an editing site. Nucleic Acids Research, 50(19), 10857–10868. 10.1093/nar/gkac897. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Eggington JM, Greene T, & Bass BL (2011). Predicting sites of ADAR editing in double-stranded RNA. Nature Communications, 2(319), 10.1038/ncomms1324. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Gholamalipour Y, Karunanayake Mudiyanselage A, & Martin CT (2018). 3’ End additions by T7 RNA polymerase are RNA self-templated, distributive and diverse in character—RNA-Seq analyses. Nucleic Acids Research, 46(18), 9253–9263. 10.1093/nar/gky796. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Huang J, Brieba LG, & Sousa R (2000). Misincorporation by Wild-Type and mutant T7 RNA polymerases: Identification of interactions that reduce misincorporation rates by stabilizing the catalytically incompetent open conformation. Biochemistry, (38), 38. 10.1021/bi000579d. [DOI] [PubMed] [Google Scholar]
  10. Hundley HA, & Bass BL (2010). ADAR editing in double-stranded UTRs and other noncoding RNA sequences. Trends in Biochemical Sciences, 35(7), 377–383. 10.1016/j.tibs.2010.02.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Jacobsen CS, Salvador P, Yung JF, Kragness S, Mendoza HG, Mandel G, & Beal PA (2023). Library screening reveals sequence motifs that enable ADAR2 editing at recalcitrant sites. ACS Chemical Biology, 18(10), 2188–2199. 10.1021/acschembio.3c00107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Katrekar D, Chen G, Meluzzi D, Ganesh A, Worlikar A, Shih YR, ... Mali P (2019). In vivo RNA editing of point mutations via RNA-guided adenosine deaminase. Nature Methods, 16, 239–242. 10.1038/s41592-019-0323-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Kebschull JM, & Zador AM (2015). Sources of PCR-induced distortions in high-throughput sequencing datasets. Nucleic Acids Research, 43(21), e143. 10.1101/008375. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Khosravi HM, & Jantsch MF (2021). Site-directed RNA editing: Recent advances and open challenges. RNA Biology, 18(sup1), 41–50. 10.1080/15476286.2021.1983288. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Lehmann KA, & Bass BL (1999). The importance of internal loops within RNA substrates of ADAR. Journal of Molecular Biology, 291, 1–13. 10.1006/jmbi.1999.2914. [DOI] [PubMed] [Google Scholar]
  16. Luas JK, Gruenke PR, & Burke DH (2023). Minimizing amplification bias during reverse transcription for in vitro selections. RNA (New York, N. Y.), 29, 1301–1315. 10.1261/rna.079650.123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Macbeth MR, & Bass BL (2007). Large-scale overexpression and purification of ADARs from Saccharomyces cerevisiae for biophysical and biochemical studies. Methods in Enzymology, 424, 319–331. 10.1016/S0076-6879(07)24015-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Matos VJ, Jacobs O, Ouye R, Mozumder S, Salvador PJ, Fink KD, & Beal PA (2024). Site-specific regulation of RNA editing with ribose-modified nucleoside analogs in ADAR guide strands. Nucleic Acids Research, 52(12), 6733–6747. 10.1093/nar/gkae461. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Matthews MM, Thomas JM, Zheng Y, Tran K, Phelps KJ, Scott AI, ... Beal PA (2016). Structures of human ADAR2 bound to dsRNA reveal base-flipping mechanism and basis for site selectivity. Nature Structural and Molecular Biology, 23(5), 426–433. 10.1038/nsmb.3203. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Milligan JF, Groebe DR, Wherell GW, & Uhlenbeck OC (1987). Oligonucleotide synthesis using T7 polymerase and synthetic DNA templates. Nucleic Acids Research, 15(21), 8783–8798. 10.1093/nar/15.21.8783. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Minshall N, & Git A (2020). Enzyme- and gene-specific biases in reverse transcription of RNA raise concerns for evaluating gene expression. Scientific Reports, 10, 8151. 10.1038/s41598-020-65005-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Monian P, Shivalila C, Lu G, et al. (2022). Endogenous ADAR-mediated RNA editing in non-human primates using stereopure chemically modified oligonucleotides. Nature Biotechnology, 40, 1093–1102. 10.1038/s41587-022-01225-1. [DOI] [PubMed] [Google Scholar]
  23. Montiel-Gonzalez MF, Vallecillo-Viejo I, Yudowski GA, & Rosenthal JJC (2013). Correction of mutations within the cystic fibrosis transmembrane conductance regulator by site-directed RNA editing. Proceedings of the National Academy of Science, 110, 18285–18290. 10.1073/pnas.1306243110. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Montiel-Gonzalez MF, Vallecillo-Viejo I, & Rosenthal JJC (2016). An efficient system for selectively altering genetic information within mRNAs. Nucleic Acids Research, 44, 157. 10.1093/nar/gkw738. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Montiel-Gonzalez MF, Diaz Quiroz JF, & Rosenthal JJC (2019). Current strategies for site-directed RNA editing using ADARs. Methods (San Diego, Calif.), 156, 16–54. 10.1016/j.ymeth.2018.11.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Oshlack A, & Wakefield MJ (2009). Transcript length bias in RNA-seq data confounds systems biology. Biology Direct, 4, 140. 10.1186/1745-6150-4-14. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Özgür S, & Orman M (2023). Application of deep learning technique in next generation sequence experiments. Journal of Big Data, 10, 160. 10.1186/s40537-023-00838-w. [DOI] [Google Scholar]
  28. Pfeiffer LS, & Stafforst T (2023). Precision RNA base editing with engineered and endogenous effectors. Nature Biotechnology, 41, 1526–1542. 10.1038/s41587-023-01927-0. [DOI] [PubMed] [Google Scholar]
  29. Qu L, Yi Z, Zhu S, Wang C, Cao Z, Zhou Z, ... Wei W (2019). Programmable RNA editing by recruiting endogenous ADAR using engineered RNAs. Nature Biotechnology, 37(9), 1059–1069. [DOI] [PubMed] [Google Scholar]
  30. Quiroz JFD, Siskel LD, & Rosenthal JJC (2023). Site-directed A → I RNA editing as a therapeutic tool: Moving beyond genetic mutations. RNA (New York, N. Y.), 2, 498–505. 10.1261/rna.079518.122. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Reautschnig P, Wahn N, Wettengel J, Schulz AE, Latifi N, Vogel P, ... Stafforst T (2022). CLUSTER guide RNAs enable precise and efficient RNA editing with endogenous ADAR enzymes in vivo. Nature Biotechnology, 40(5), 759–768. [DOI] [PubMed] [Google Scholar]
  32. Schmidt B, & Hildebrandt A (2021). Deep learning in next-generation sequencing. Drug Discovery Today, 26(1), 173–180 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7550123/. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Schneider MF, Wettenge J, Hoffmann PC, & Stafforst T (2014). Optimal guidernas for re-directing deaminase activity of hadar1 and hadar2 in Trans. Nucleic Acids Research, 42(10), 17. 10.1093/nar/gku272. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Sinnamon JR, Kim SY, Corson GM, Song Z, Nakai H, Adelman JP, & Mandel G (2017). Site-directed RNA repair of endogenous MeCP2 RNA in neurons. Proceeding of the National Academy of Sciences, 114, e9395–e9402. 10.1073/pnas. 1715320114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Sinnamon JR, Kim SY, Fisk JR, Jeng S, McWeeney SK, & Mandel G (2020). In vivo repair of a protein underlying a neurological disorder by programmable RNA editing. Cell Reports, 32(2), 107878. 10.1016/j.celrep.2020.107878. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Sinnamon JR, Jacobsen ME, Yung JF, & Mandel G (2022). Targeted RNA editing in brainstem alleviates respiratory dysfunction in a mouse model of Rett syndrome. PNAS, 119(33), e2206053119. 10.1073/pnas. 2206053119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Stebbins CC, Petrillo M, & Stevenson LF (2019). Immunogenicity for antisense oligonucleotides: A risk based assessment. Bioanalysis, 11, 1913–1916. 10.4155/bio-2019-0133. [DOI] [PubMed] [Google Scholar]
  38. Suter A, 2024. Wave Life Sciences announces approval of first clinical trial application for restoraation-2 trial of WVE-006 in individuals with alpha-1 antitrypsin deficiency (AATD). Wave Life Sciences. https://ir.wavelifesciences.com/news-releases/news-release-details/wave-life-sciences-announces-approval-first-clinical-trial. [Google Scholar]
  39. Verwilt J, Mestdagh P, & Vandesompele J (2023). Artifacts and biases of the reverse transcription reaction in RNA sequencing. RNA (New York, N. Y.), 29, 889–897. 10.1261/rna.079623.123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Wong BL, Mendoza HG, Jacobsen CS, & Beal PA (2024). RNA sequences that direct selective ADAR editing from a SELEX library bearing 8-azanebularine. Bioorganic & Medicinal Chemistry. 10.1016/j.bmc.2024.117700. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Wong SK, Sato S, & Lazinski DW (2001). Substrate recognition by ADAR1 and ADAR2. RNA (New York, N. Y.), 7(6), 846–858. 10.1017/s135583820101007x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Wulff BE, & Nishikura K (2010). Substitutional A-to-I RNA editing. WIREs RNA, 90–101. 10.1002/wrna.10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Yu AM, Choi YH, & Tu MJ (2020). RNA drugs and RNA targets for small molecules: principles, progress, and challenges. Pharmacology Reviews, 72(4), 862–898. 10.1124/pr.120.019554. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Zhang JX, Yordanov B, Gaunt A, et al. (2021). A deep learning model for predicting next-generation sequencing depth from DNA sequence. Nature Communications, 12, 4387. 10.1038/s41467-021-24497-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Zucha D, Androvic P, Kubista M, & Valihrach L (2020). Performance comparison of reverse transcriptases for single-cell studies. Clinical Chemistry, 66(1), 217–228. 10.1373/clinchem.2019.307835. [DOI] [PubMed] [Google Scholar]

RESOURCES