Chemical Modifications of Adenine Base Editor (ABE) mRNA Expand Its Application Scope

Posted in: Therapeutics

  • The Liu Lab’s ABE Uses an Engineered CRISPR/Cas9 System to Convert a Single “A” in DNA to “G” 
  • The Key Component is an Evolved Adenine Deaminase Fused to a ss-DNA Cas9 Nickase
  • ABE Efficiency Is Enhanced Using TriLink’s 5-Methoxyuridine-Modified CleanCap® 1 mRNA Methodology 
David R. Liu (2012 photo; D.R.L. is a consultant and co-founder of Editas Medicine, Pairwise Plants, Prime Medicine, and Beam Therapeutics, companies that use genome editing). Taken from and free to use.

This blog will focus on a publication in Nature Communications by über-famous Prof. David R. Liu (pictured here) and collaborators (Jiang et al.) that described significant improvements in CRISPR-Cas9-associated adenine base editing (ABE). This technology precisely converts a targeted A to G, in order to correct pathogenic single-nucleotide mutations in research or therapeutic settings. In addition to this work involving Liu, a widely acclaimed superstar researcher and serial entrepreneur, the Zone was mightily impressed by the fact that according to Nature Communication metrics, this Liu publication has received ~6,500 article accesses since appearing in April 2020.

The improvements to the ABE process, schematically shown here, were enabled by several advantages provided by TriLink’s 5-methoxyuridine (5moU)-modified CleanCap® 1 mRNA methodology. This proprietary technology has been characterized in a recent expert review as a “revolution” in synthetic mRNA. Details on this TriLink technology are available in the original research report published by McCaffrey and others, as well as in a 2021 protocol published by Henderson et al. 

Schematic diagram of CRISPR DNA adenine base editing (ABE) comprised of two key components: a Cas9 single-strand endonuclease for programmable guide RNA (gRNA)-directed DNA binding and nicking (Cas9 nickase, nCas9), and an Escherichia coli tRNA adenosine deaminase evolved to accept adenine in DNA for deamination to guanosine to generate inosine, which has the same base pairing preferences as a guanosine in DNA. This leads to transformation of A→G and overall conversion of a T·A to C·G base pair. Taken from Kantor et al. and free to use. © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license

Both McCaffrey and Henderson are among the co-authors of the featured ABE article by Jiang et al., which is the current version of the Liu group’s seminal publication (Gaudelli et al.) on the ABE approach first reported in 2017 in Nature and has absolutely amazing metrics: more than 45,000 article accesses to date. 

Along with related innovative base editing research, Prof. Liu will present an enlightening overview of all of this in an upcoming GEN Webinar on April 29th titled Precision Genome Editing without Double-Strand Breaks, proudly sponsored by TriLink.


According to the original report on ABE by Gaudelli et al., G·C-to-A·T base pair substitution accounts for approximately half of known pathogenic single nucleotide mutations in humans. The ABE system, which is constructed by fusing an artificially evolved adenine deaminase to a mutated Cas9 that is a single-strand DNA nickase, can precisely and permanently convert T·A to C·G (see above Scheme) with the guidance of a target-specific guide RNA (gRNA) without creating a double-strand DNA break and without requiring an exogenous DNA repair donor. ABE therefore represents a useful tool in the modeling or treatment of genetic diseases associated with single nucleotide mutations, and many ABE variants have already been widely studied and applied.

A 3D model of adeno-associated virus (AAV) serotype 1, built using data of viral molecular structure from Protein Data Bank (PDB 3NG9).

However, note Jiang et al., effective base editing requires high cellular expression of ABE agents, which limits the application of ABE-encoding DNA plasmids in cells that are difficult to transfect. They add that, while the therapeutic potential of ABE has been demonstrated by delivery of DNA-encoded base editors to adult animal disease models via plasmids or adeno-associated virus (AAVs), these approaches raise the potential for DNA integration or off-target effects due to long-term exposure to the gene-editing machinery, thus hindering their clinical relevance. 

Consequently, according to Jiang et al., ‘[d>

eveloping non-DNA-encoded base editor and non-viral delivery methods may facilitate broader application of ABE.” Base editing, first demonstrated in 2016 by Liu and coworkers for cytidine deamination to uridine (i.e., C-to-T following DNA replication of U as T), led to microinjection of cytidine base editor (CBE) mRNA and in vitro transcribed guide RNA for effective base editing in mouse embryo or pig oocyte. However, noted Jiang et al., “the editing efficiency mediated by an mRNA-encoded ABE system has not been studied in somatic cells, and its application potential and delivery method have not been addressed.”

3D illustration of a simplified empty LNP structure.

To investigate this possibility, Jiang et al. engineered an mRNA-encoded ABE system by introducing various chemical modifications to both ABE mRNA and gRNA. As outlined in the following sections, at some genomic sites, the optimized base editing system exhibits higher editing efficiency compared to that provided by a DNA-encoded system. Furthermore, Jiang et al. demonstrate that this RNA-based system mediates robust editing in hard-to-transfect cystic fibrosis bronchial epithelial cells. Moreover, by encapsulating chemically modified ABE mRNA and gRNA into liposome-like lipid nanoparticles (LNPs), they successfully deliver the mRNA-encoded ABE into the liver of tyrosinemia I mice, correct the disease mutation, and rescue the phenotype. 

As reviewed elsewhere, hereditary tyrosinemia I (HT I) is a genetic disorder of tyrosine metabolism characterized by progressive liver damage starting in infancy, as well as by a high risk for hepatocellular carcinoma. HT I is due to mutations in the fumarylacetoacetate hydrolase (Fah) gene, which encodes the last enzyme in the tyrosine catabolic pathway. Disturbances in tyrosine metabolism lead to increased levels of succinylacetone and succinylacetoacetate. The mechanisms causing liver failure, cirrhosis, renal tubular dysfunction, and hepatocarcinoma are however still unknown.

Methods and Results

Uridine-depleted ABE Cas9 mRNA modified with 5-moU: In previously published work, Liu and collaborators (Song et al.) engineered a codon-optimized variant of ABE (termed RA6.3) that recognizes an “NGG” (N = A/G/C/T) protospacer adjacent motif (PAM) sequence with improved  editing efficiency in human embryonic kidney (HEK293T) cells. Hydrodynamically-injected plasmids delivered RA6.3 (depicted here) and gRNA to mouse liver and corrected a splice-site mutation of the Fah gene.

Schematic structure of RA6.3. NLS: nuclear localization signal. TadA is a tRNA adenine deaminase. TadA* denotes the evolved TadA version 6.3 (Gaudelli et al.) that deaminates adenine in DNA. Taken from Jiang et al. and free to use (Open Access under a Creative Commons Attribution 4.0 International License).

However, when the researchers tested non-viral, i.e., LNP-mediated delivery of RA6.3 encoded by unmodified mRNA in vivo, the editing efficiency was 3-fold lower than that of plasmid-delivered RA6.3 in Tyrosinemia I mice. Based on these earlier results, Jiang et al. speculated that the lower editing rate may be due to instability of unmodified RA6.3 mRNA, and subsequently, poor expression in cells. To increase cellular expression, Jiang et al. set out to optimize the chemical composition of ABE mRNA.

Based on the original report by McCaffrey and collaborators, use of synonymous codons, replacement of all of remaining uridines with 5moU, and use of Cap 1 allow uridine depletion to increase Cas9 mRNA activity. In light of this, Jiang et al. engineered three versions of 5’ capped RA6.3 mRNA: unmodified, uridine-depleted, and 5moU-modified, all depicted here.   

Diagrams of ABE RA6.3 mRNAs with sequence-optimization (uridine-depletion) and chemical modification (5-methoxyuridine). Red: “A”; Yellow: “U”; Green: “G”; Blue: “C”. Taken from Jiang et al. and free to use (Open Access under a Creative Commons Attribution 4.0 International License).

Briefly, uridine depletion was carried out using the “optimize codons” tool in Geneious version R8.0.5, and mRNAs were synthesized by T7 RNA polymerase in vitro transcription (IVT). Using the CleanCap® Reagent AG, unmodified or 5moU modified (depicted here) mRNAs were co-transcriptionally capped to produce Cap 1 mRNAs, as shown here. IVT was performed with 0.025 μg/μL standard or uridine-depleted transcription template, 4mM CleanCap® Reagent AG trimer, and 5 mM each of ATP, CTP, GTP, and UTP or 5moUTP. 

5-Methoxyuridine (5moU) unit in modRNA, derived from IVT using 5-methoxyuridine-5'-triphosphate and T7 RNA Polymerase. Incorporation of 5moU can reduce the immunogenicity of the resulting mRNA. Taken from TriLink BioTechnologies.
CleanCap® AG analog for co-transcriptional capping with T7 RNA polymerase. (A) Structure of CleanCap® AG trimer. (B) T7 RNA polymerase promoter sequence (underlined) with initiation sequence required for CleanCap AG®. Arrow indicates transcription start site with nucleotide positions shown above. Taken from Henderson et al. Current Protocols, 1, e39 with permission. 

After transient transfection into HEK293T cells, only 5moU mRNA-encoding RA6.3 (hereafter termed 5moU-6.3) yielded stable protein expression comparable to chemically modified Cas9 mRNA (5meC/pseudouridine) obtained from TriLink Cas9 mRNA (Fig. 1b). The expression dynamics of 5moU-6.3 were also similar to Cas9 mRNA: detectable at 6 h post-transfection, reaching peak expression after 1 day, and degraded after 2 days. Thus, Jiang et al. used 5moU-6.3 to test editing efficiency in the following experiments.

Moderately modified gRNA mediates robust editing: The guide RNA (gRNA) of the CRISPR-Cas9-associated editing system is comprised of a CRISPR RNA (crRNA) and a transactivating crRNA (tracrRNA), as depicted here for generalized gene editing. Cas9 protein is not shown, and the dashed line indicates covalent bonding of crRNA and tracrRNA to give a single gRNA (sgRNA aka gRNA) construct as an alternative for targeting DNA.

Generalized diagram for CRISPR/Cas9 genome editing.

After electroporating HEK293T cells with 5moU-6.3 mRNA and unmodified tracrRNA/crRNA, minimal editing (<1%) was observed at a genomic site where DNA-encoded RA6.3 and gRNA mediated an average A-to-G conversion rate of 19.5%, as described before (Song et al.). To improve RNA-encoded base editing efficiency and based on previous reports indicating that chemical modifications improve gRNA stability, Jiang et al. designed “moderately modified” or “heavily modified” tracrRNA and crRNA (see Supplementary Fig. 1d), then co-electroporating these constructs with 5moU-6.3 to compare editing efficiencies.  

Moderately modified tracrRNA/crRNA (i.e., 2’-O-methyl 3’-phosphorothioate modifications at the first and last three nucleotides) conferred the highest editing efficiency (see Supplementary Fig. 1e, f), and was comparable to gRNA expressed from plasmid. Because plasmid-expressed gRNA is sgRNA, the researchers then designed moderately modified sgRNA constructs (see Supplementary Fig. 1g). 5moU-6.3 mRNA and these modified sgRNAs could result in A-to-G conversion rates comparable to those provided by RA6.3 and sgRNA expressed from plasmids. Jiang et al. also tested the editing efficiency of 5moU-6.3 and moderately modified sgRNA at two genomic sites in which DNA-encoded RA6.3 mediated limited conversion rates. The RNA-encoded ABE system showed significantly higher editing rates at all “A” sites within the editing window.

Next, Jiang et al. compared the editing efficiency of unmodified and 5moU-modified ABE mRNA when delivered with moderately modified sgRNA. They first measured the A-to-G conversion rate at the same genomic site mentioned above at different concentrations (ranging from 0.0015 to 0.5 μg) of unmodified-6.3 or 5moU-6.3. It was found that at lower dosages (less than 0.015 μg), 5moU-6.3 shows ~1.5-fold higher editing efficiency compared to unmodified mRNA. Similarly, at the two other genomic sites mentioned above, 5moU-6.3 mediated substantially higher A-to-G conversion rates than those mediated by unmodified mRNA, indicating robust RNA-encoded ABE. 

To compare the off-target effects of DNA vs. modified RNA-encoded ABE systems, Jiang et al. analyzed A-to-G conversion rates at the top known off-target loci for the two gRNAs mentioned above. High-throughput “deep” sequencing (Illumina) data show that, overall, base editing rates at the off-target sites are low (<0.2% in all groups) (see Supplementary Fig. 3a, b). A significant increase in A-to-G conversion rate was only detected at 2 of the 14 “A” sites in 5moU-6.3-treated cells compared to control or DNA-encoded ABE-treated cells, thus indicating improved RNA-encoded vs. DNA-encoded ABE.

Structural model of the CFTR protein. Taken from and free to use.

ABE corrects a nonsense mutation in a cystic fibrosis model: Next, Jiang et al. investigated whether their engineered RNA-encoded ABEsystem can correct a pathogenic single nucleotide mutation in a cystic fibrosis (CF) cell model. According to Jiang et al., ~10% of CF patientscarry cystic fibrosis transmembrane conductance regulator (CFTR, shown here) nonsense mutations that cannot be treated with anyFDA-approved CFTR modulators, and around half of thesemutations are correctable by ABE. Thesecond most common CFTR nonsense mutation, termed W1282X,is caused by a G→A mutation in exon 23 that produces minimalamount of functional CFTR protein and abolishes its chloride ion (Cl) transportactivity. A previous report showed that a human bronchialepithelial cell line (abbreviated 16HBEge)homozygous for the W1282X mutation, recapitulates the phenotypeof primary W1282X CF cells, making it a good cell model for the study of howto correct this CFTR mutation. 

Because electroporated mRNA expressed well in this cell line, Jiang et al. could test their RNA-encoded ABE system for correcting the CFTR W1282X mutation. Specifically, RA6.3 could correct this mutation by recognizing the “TGG” PAM sequence, wherein the target “A” site falls at position 9 (A9) of a protospacer that RA6.3 could use to correct the mutation. These researchers electroporated modified sgRNA with either unmodified-6.3 or 5moU-6.3 into 16HBEge cells to compare editing efficiency of unmodified vs. 5-moU-modified ABE mRNA. 

Briefly, it was found that 5moU-6.3 achieved a significantly higher A-to-G conversion rate (26.4 ± 7.40%) at the A9 target site compared to unmodified-6.3 (13.1 ± 0.51%). Furthermore, 5moU-6.3 restored full-length CFTR protein expression to ~10% of the level in wild-type cells, based on Western blots. Notably, the editing efficiency at the bystander “A” site at position 5 (45.1 ± 5.66%) was higher than the target A9 site, and changed codon 1281 from glutamine (Q) to arginine (R). However, a series of experiments to investigate whether this amino acid alteration affects CFTR function demonstrated that the bystander Q1281R mutation does not affect CFTR function. It was therefore concluded that RA6.3 can correct a CF mutation, and the RNA-encoded ABE system can mediate robust editing in hard-to-transfect cells.


ABE corrects a splice-site mutation in a Tyrosinemia I mouse model: Finally, Jiang et al. tested the non-viral delivery of ABE to the liver in a mouse model of Tyrosinemia I harboring a homozygous G·C-to-A·T point mutation in the last nucleotide of exon 8 in the Fah gene, which causes exon 8 skipping, FAH protein deficiency, and liver damage. To maintain body weight and survival, these mice are given water supplemented with NTBC [2-(2-nitro-4-trifluoromethylbenzoyl)-1,3-cyclohexanedione>

, a tyrosine catabolic pathway inhibitor.

In principle, RA6.3 can use a protospacer that places the target “A” at position 9 to correct this causative mutation. Hepatocytes with corrected FAH protein will gain growth advantage and eventually repopulate the liver. Jiang et al. used LNPs to separately encapsulate 5moU-6.3 and modified sgRNA, and delivered these LNPs via tail vein injection into adult mouse liver at four dosages, based on their previously reported studies (Song et al.). One week after the last injection, they replaced NTBC-supplemented water with normal water to allow for the repopulation of corrected cells.

Briefly, treated mice maintained their body weight, while the untreated mice rapidly lost their body weight, suggesting restoration of FAH function in LNP-treated mice. Jiang et al. observed widespread FAH-positive patches in mouse liver by staining with a FAH-specific antibody, and the protein restoration rate was comparable to that in mice treated with DNA plasmid-delivered RA6.3 via hydrodynamic injection. Next, they investigated efficiency of Fah gene correction. At the transcriptome level, RT-PCR results from liver tissues of LNP-treated mice revealed that the majority of Fah mRNA was now properly spliced. At the DNA level, they observed an A→G correction rate of 12.5 ± 2.7% at the target site in the liver tissues of LNP-treated mice, which is comparable to their previously reported plasmid-delivered ABE system (9.5 ± 4.0%) (Song et al.).  

Notably, there is a bystander site (A6) in the editing window. The A→G conversion at this site changes a serine codon into alanine (S235A) in the FAH enzyme. Because S235 is near the FAH enzyme active site, A→G conversion at the A6 site will not rescue the splicing defect but may affect enzyme activity. Surprisingly, say Jiang et al., at this bystander site, they observed a significantly lower editing rate (0.096 ± 0.032%) in LNP-treated mice compared to their previously reported rate (1.9 ± 0.9%) (Song et al.) (P = 0.0006) using a plasmid, suggesting that the relatively short half-life of ABE mRNA might minimize bystander conversions in vivo


Jiang et al. concluded that “[t>

his LNP-mediated non-viral delivery of modified RNA-encoded ABE, unlike hydrodynamic injection of plasmid DNA, provides a clinically-relevant therapeutic method for genetic diseases caused by single nucleotide mutations.”

Concluding Comments

By applying various chemical modifications to ABE mRNA and RNA, Jiang et al. successfully engineered an RNA-encoded ABE system. Although previous reports by others show that unmodified cytidine base editor (CBE) mRNA and gRNA could effectively edit embryos and oocytes, the above findings by Jiang et al. show that unmodified ABE mRNA does not effectively express in HEK293T cells, and unmodified gRNA cannot mediate efficient editing in somatic cell culture. These researchers suggested that this might be because unmodified mRNA and gRNA are relatively unstable and will quickly undergo degradation after being delivered to cell culture.

Jiang et al. further demonstrated that chemical modifications are essential for RNA-encoded ABE to mediate efficient editing. They showed that uridine depletion and 5-moU modification is critical for the stable expression of ABE mRNA. Furthermore, Jiang et al. “expect this optimized RNA system will be applicable for other CRISPR-associated editors, e.g., prime editors reported by Liu et al. Importantly, Jiang et al. did not observe significant off-target effects using their modified mRNA-encoded ABE at top-known off-target sites for gRNAs. However, they add that “future work should perform unbiased screens to detect off-target editing by modified ABE mRNA at the whole genome and transcriptome levels.”


Jiang et al. observe that, unlike the well-established CRISPR-Cas9 field, there are very few studies that use protein or RNA-encoded base editing to correct diseases. The findings outlined above, they add, “provide the first report on delivery of RNA-encoded ABE to effectively correct disease-causing point mutations in vitro and in vivo.” The successful correction of a currently untreatable CF mutation demonstrates the potential of ABE as a new gene therapy method for CF treatment. 

Specifically, LNP-based non-viral delivery of ABE successfully corrects a Fah gene point mutation in Tyrosinemia I mice, providing a clinically relevant method for the treatment of genetic diseases. Corrected cells with normal FAH function have a growth advantage. They outgrow the non-corrected cells, which magnifies the therapeutic effect. 

Finally, Jiang et al. conclude that, to provide a delivery method suitable for treating a broad array of diseases, “future work should optimize delivery dosage and nanoparticle formulations for ABE encapsulation to maximize initial editing efficiency. In summary, [these>

optimized ABE mRNA and gRNA reagents unlock new therapeutic possibilities by using ABE.”

In conclusion, it is important to mention that on January 6th, 2021, Liu and collaborators (Koblan et al.) reported in Nature that lentiviral delivery of ABE can be used to directly correct the pathogenic point mutation in LMNA, the gene that encodes nuclear lamin A associated with  Hutchinson–Gilford progeria syndrome (HGPS aka progeria), in ~90% of cultured fibroblasts derived from children with progeria and in a mouse model of HGPS. These findings demonstrate the potential of in vivo ABE as a possible treatment for HGPS, as the therapy would directly correct the root cause of the disease. Koblan et al. suggested that non-viral LNP-mediated delivery of RNA-encoded ABE may ameliorate future treatment of HGPS. The Zone is mightily impressed by the aforementioned improvements to ABE that were enabled by TriLink’s 5-methoxyuridine (5moU)-modified CleanCap® 1 mRNA methodology, and we look forward to posting follow-up blogs on notable future ABE publications that use this technique. 


What do you think?

Your comments are welcomed, as usual.

PS: Remember to mark your calendar for Prof. Liu’s upcoming GEN Webinar on April 29th, titled Precision Genome Editing without Double-Strand Breaks, proudly sponsored by TriLink.

Please feel free to share this blog with your colleagues or on social media. 

2 years ago
163 view(s)
Did you like this post?