Genome, Epigenome, and Transcriptome Editing via the Chemical Modification of Nucleobases in Living Cells

Abstract

Base editors are tools that chemically modify the nucleobases of DNA and RNA in a programmable manner, allowing for genome, epigenome, and transcriptome editing in live cells. These tools can be used to introduce specific base transitions in DNA or RNA, manipulate methylation patterns in the epigenome, and create genetically encoded libraries in target genes. These various functions can be used to modulate every aspect of the central dogma. The efficiency and precision of base editors makes them useful in both basic research and in the development of new therapies. Here we describe currently available base editors and the ways they can be used to better understand and manipulate different aspects of the central dogma.

Graphical Abstract

Introduction

The development of tools that directly chemically modify the nucleobases of DNA and RNA (via enzymatic methylation, deamination, or demethylation) in living cells has opened the door for studying and manipulating the components of the central dogma. These tools have been enabled by an increasing understanding of CRISPR systems and their ability to recognize and bind to specific dsDNA (for Cas9 and Cas12 enzymes)¹ and ssRNA (for Cas13 enzymes)² sequences inside living cells. In these systems, a guide RNA (gRNA) is bound by a Cas enzyme to form a ribonucleoprotein (RNP) complex. The RNP recognizes a target locus (the protospacer) through base pairing between the gRNA and the target nucleic acid sequence. When the target is DNA, this process separates the two DNA strands, creating a DNA-RNA heteroduplex with a small single-stranded DNA (ssDNA) region at the target locus (known as an R-loop).³ Once bound to its target dsDNA sequence, the Cas enzyme cleaves both phosphodiester backbones of the target nucleic acid using either a single (in the case of Cas12 enzymes)^4–6 or two distinct (in the case of Cas9)¹ endonuclease domains. If the target is ssRNA, the Cas13 enzyme becomes activated upon target binding, and cleaves the phosphodiester backbone of both the target sequence and any neighboring ssRNA molecules in a promiscuous manner.^7–9 This RNA-guided endonuclease activity has been utilized extensively for genome editing and RNA degradation purposes.¹⁰

The catalytic activity of Cas enzymes can be inactivated by introducing specific mutations into the endonuclease domains to create catalytically dead Cas variants (dCas) that use a gRNA to bind a target DNA or RNA sequence without cleaving the phosphodiester backbone.¹¹ A diverse set of genome, epigenome, and transcriptome editing tools (collectively referred to here as base editors) have been developed from these dCas proteins by physically attaching them to various nucleobase modifying enzymes.¹² In this way, the chemical activity of these enzymes can be confined to specific genomic and transcriptomic loci where they chemically modify a canonical nucleobase into either a noncanonical DNA base or a naturally occurring modified base. As such, different base editing tools allow researchers to modify and study the effects of nucleic acid primary sequences, chromatin organization, RNA activity and stability, and DNA damage. In this perspective, we discuss the various base editor tools available for directly chemically modifying target nucleobases in the genome and transcriptome, and their potential for manipulating the composition and expression of target genes.

Cytosine deamination

Base editing tools were first created by fusing cytidine deaminase enzymes to dCas9.^13,14 These deaminases recognize cytosine in a ssDNA context and deaminate the nucleobase to form uracil, which has the base pairing properties of thymine (Figure 1). Due to the substrate requirement of ssDNA, the deaminase activity of the tethered enzyme is confined to only the single-stranded portion of the dCas9:gRNA:DNA R-loop. This strategy results in an editing “window” of only five to nine nucleotides (depending on the Cas9 variant used) within the protospacer.^13,15,16

Figure 1.

Nucleobase chemistries facilitated by current base editors. Cytosines can be methylated to 5-methylcytosines by DNMT3A-derived base editors. Likewise, 5-methylcytosines can be demethylated by TET1-containing base editors. These two types of tools can be used to control epigenetic patterns. Cytosine and adenine can be deaminated to uracil and inosine, respectively. These chemical transformations are catalyzed by base editors comprised of APOBEC or AID in the case of uracil, and TadA or ADAR in the case of inosine in DNA or RNA, respectively. Uracil is further converted to thymine and inosine is further converted to guanine by cellular replication or repair processes due to the new hydrogen-bonding properties of these base editing intermediates. Uracil is efficiently excised by the base excision repair protein UDG to form an abasic site, which is mutagenized to all four canonical DNA bases (dashed lines). The details of this process are still relatively unknown and the outcome is largely uncontrollable and unpredictable.

Several tools have since been created that link a cytidine deaminase to a DNA targeting enzyme such as dCas9^13,14,17,18 or dCas12a.¹⁵ Though each tool may have slight variations, they each use the Cas protein to bring the cytidine deaminase to a target locus and convert cytosines to uracils. The final outcome following uracil incorporation, however, depends on the tool being used. Mutagenesis tools such as Targeted AID-mediated mutagenesis (TAM)¹⁹ and CRISPR-X¹⁸ rely on the cell’s efficient excision of uracil from genomic DNA by the base excision repair pathway (BER). Excision of uracil by uracil-DNA glycosylase (UDG) creates an abasic site which can be permanently converted to any of the four canonical DNA bases through an as-yet-unknown mechanism that potentially involves error-prone DNA polymerases (Figure 1).²⁰

A particularly intriguing feature of these mutagenesis tools is their ability to generate genetically-encoded libraries. Within living cells, a library of gRNAs can be used to target these tools across an entire gene to introduce mutations. When combined with an appropriate selection or screen, these libraries can be used to identify mutations that give rise to specific phenotypes. Both TAM and CRISPR-X have been used to identify protein variants that give rise to chemotherapeutic resistance.^18,19 The ability to rapidly identify mutations that cause drug resistance (or another specific phenotype of interest) in a cellular context is a powerful tool that holds great promise in the areas of personalized medicine and reverse genetics.

Cytidine deaminase-derived base editing tools such as BE1-4^13,21 and Target-AID¹⁴ can also be used to create predictable C•G to C•A mutations by manipulating cellular DNA repair pathways. Specifically, the bacteriophage protein uracil glycosylase inhibitor (UGI) can be physically linked²² to or co-expressed²³ with the base editor. UGI protects the uracil intermediate from excision by reversibly binding to UDG²⁴ and significantly decreases C•G to non-T•A mutation rates.²¹ Alternatively, the uracil intermediate can be avoided entirely by directly deaminating a methylated cytosine, which results in a direct conversion to thymine.²⁵ To manipulate DNA repair pathways, the catalytic activity of a single endonuclease domain in dCas9 can be restored to create a Cas9 nickase (Cas9n).²⁶ During base editing, Cas9n creates a nick in the DNA strand opposite the uracil. This nick marks the strand as “newly synthesized”, resulting in the mismatch repair pathway (MMR) repairing the U•G mismatch using the uracil-containing strand as a template.²⁷ By manipulating the MMR pathway in this way, the guanine of the original C•G base pair is replaced with an adenine, solidifying the desired edit before uracil can be excised. These features can be combined (as in BE3, BE4, and Target-AID) to maximize C•G to T•A base editing efficiency.

The ability to precisely introduce single or multiple uracil lesions at a pre-defined location of the genome of living cells makes base editors valuable tools for both therapeutic and research purposes. Beyond using cytidine deaminase base editors to correct disease-causing point mutations and create genetically-encoded libraries, these tools have also unearthed surprising new properties of uracil repair. For example, it was discovered that single G•U lesions are repaired through a different, more error-prone mechanism than multiple clustered G•U lesions.²¹ Though the mechanism of this differential processing is still not fully understood, its elucidation will likely be aided by using cytidine deaminase base editors as precision DNA damaging tools. Likewise, base editors capable of installing other types of DNA damage will help further our understanding of other mechanisms of DNA repair.

Cytosine methylation and 5-methylcytosine oxidative demethylation

Methylation of cytosines at the 5 position results in 5-methylcytosine (5mC), a naturally occurring modified DNA base (Figure 1). The presence of 5mC in promoter regions (where it occurs mostly at CpG sites) has a significant influence on gene expression at the transcriptional level. As a general rule, hypermethylation of these regions results in gene silencing.^28,29 The natural methylation of cytosines and demethylation of 5mC’s are facilitated by DNA methyltransferase (DNMT) and ten-eleven translocation methylcytosine dioxygenase (TET) enzymes, respectively (Figure 1). Modulation of 5mC levels is particularly essential during development as shifting methylation levels control cell differentiation.³⁰ Inappropriate silencing due to hypermethylation can have devastating effects, as observed in several cancers and fragile X syndrome.³¹ Therefore, tools that manipulate methylation levels at target regions are greatly beneficial for studying and developing treatments for a variety of diseases.

Two types of base editors have been created that allow researchers to modulate 5mC levels in target regions of the genome.^32–39 A DNMT3A-dCas9 fusion protein was developed as a tool to increase methylation levels. DNMT3A recognizes cytosines in dsDNA and uses S-Adenosyl methionine (SAM) as a methyl donor to create 5mC (Figure 1). To create a tool that decreases methylation levels, the enzymatic domain of TET1 was fused to dCas9. TET1 recognizes 5mC’s and uses α-ketoglutarate to oxidize the methyl group to form 5-hydroxymethylcytosine as the first step of cytosine demethylation. 5-hydroxymethylcytosine is then further oxidized and subsequently excised by thymine DNA glycosylase to ultimately yield unmethylated cytosine (Figure 1).⁴⁰ As the natural substrates of these enzymes are dsDNA, the activity of these epigenome editors is not constrained within the ssDNA region of the R-loop. This allows the enzymes to access and modify a large window of nucleobases surrounding the binding site of the editor. This window is further widened due to the tight coiling of chromatin, which can allow the enzyme to access genomic DNA that is potentially thousands of base pairs away from its target site, but spatially very close.

These epigenome editors are quite valuable for studying the effects of methylation in untranslated regions (UTR) of the genome. For example, they have been instrumental in understanding methylation effects in fragile X syndrome.⁴¹ This genetic disease is caused by a CGG trinucleotide repeat expansion in the 5’ UTR of the FMR1 gene, which codes for fragile X mental retardation protein (FMRP).⁴² Hypermethylation of this UTR leads to silencing of FMR1 and loss of FMRP expression. By targeting the dCas9-TET1 editor to this region, demethylation was observed and FMRP expression was recovered, leading to alleviation of the phenotype in post-mitotic neurons.

Adenine deamination

Cytosine is unique among the DNA bases as it is modified by several naturally occurring enzymes. While this facilitated the creation of the first base editors, the development of base editors that modify DNA bases other than cytosine is complicated by the lack of natural enzymes that can be repurposed to perform this chemistry. Fortunately, RNA nucleobases are extensively post-transcriptionally modified by natural enzymes.⁴³ By evolving one of these RNA-modifying enzymes to accept ssDNA as a substrate, an adenine base editor (ABE) that deaminates target adenines to inosines was engineered.¹⁷ Adenosine deamination substitutes the amino group for a keto group and alters the base pairing properties of the nucleobase to match those of guanine.⁴⁴ As such, ABE catalyzes an overall A•C to G•C edit in genomic DNA at a locus programmed by the gRNA (Figure 1).

Intriguingly, A•C to G•C editing with ABE exhibits far lower rates of random mutagenesis than uracil-derived base editors, suggesting significant differences in cellular excision efficiencies of inosine (by the DNA-3-methyladenine glycosylase enzyme, AAG, and homologs)^45,46 as compared to uracil. As spontaneous adenosine deamination is 50-times slower than cytosine deamination,⁴⁷ the cellular repair machinery for uracil must be more efficient than that of inosine to preserve the integrity of the genome. Consequently, ABE-installed inosines are excised less efficiently than uracils, resulting in a more consistent mutagenic outcome for ABE. This highlights another interesting feature of DNA repair discovered through the use of base editors, and suggests that less common types of DNA damage can be used to create genome editing tools with more predictable outcomes.

While making modifications to DNA nucleobases can create a permanent mutation or modification in a cell, transiently mutating or modifying the transcriptome may be desirable in certain situations. RNA Base editing can be achieved using SNAP-ADAR,⁴⁸ λN-ADAR,⁴⁹ and RNA Editing for Programmable A to I Replacement (REPAIR).⁵⁰ All three of these technologies rely on a fusion complex of the adenosine deaminase acting on RNA 2 (ADAR2) enzyme with a gRNA molecule (SNAP-ADAR and λN-ADAR) or the dCas13b, which uses a gRNA to bind to target RNA sequences. The gRNA is designed to base pair with a target mRNA sequence and form a C•A mismatch (with the adenosine on the mRNA strand). ADAR2 deaminates the mismatched adenosine to inosine more efficiently⁵¹ than any well-matched adenosines in the protospacer, providing single nucleotide resolution. The resulting inosine exhibits the base pairing properties of guanine during translation, and thus RNA base editing can be used for the transient expression of mutant proteins.

Along with transient base editing, SNAP-ADAR, λN-ADAR, and REPAIR open up the opportunity for better understanding the roles of RNA modifications. RNA modifications occur across all domains of life and affect the activity, localization, and stability of RNA. Over 100 types of modifications have been identified, and they have been found to exist in all types of RNAs.⁴³ Tools that allow researchers to install these modifications throughout the transcriptome will significantly aid in illuminating their functions and mechanisms.

Conclusions and Future Outlook

Most importantly, base editing has been shown to work in a variety of in vivo contexts. Successful C•G to T•A base editing has been accomplished in plants, including rice, wheat, maize, and tomato,^52–54 as well as a variety of animals including silkworms,⁵⁵ zebrafish,^56,57 mice,^56,58 and human embryos. ABE has been used in rats,⁵⁹ mice,⁶⁰ rice,^61–63 and wheat.⁶¹ Furthermore, base editing has been performed in post-mitotic sensory cells, showing that this technology does not require cellular replication.⁶⁴ These experiments show that base editing is compatible with a large variety of cell types and organisms, making it valuable in many different areas of research.

ABE and cytosine deamination base editors can be used to study protein mutants and disease-associated intronic and non-coding mutations in an endogenous manner. Current strategies to study protein mutants involve knocking-out or silencing the corresponding gene, followed by overexpressing a variant to observe phenotypic effects. This method often results in protein expression levels that are significantly different from the endogenous system. Additionally, this strategy is difficult to apply to the study of point mutations that occur in intronic and non-coding regions of the genome. While using wtCas9 and a donor template can be used to edit endogenous genomic loci,⁶⁵ base editors can introduce point mutations with higher efficiency and fewer genome editing byproducts than DSB-reliant methods. This allows for the more rapid study of multiple variants in parallel, and presents the potential to efficiently introduce mutations at different sites throughout the genome in a given cell. Furthermore, base editors can be used to knock out genes of interest by introducing early stop codons via a method termed CRISPR-STOP.⁶⁶ This method has advantages over traditional, DSB-reliant methods for gene knock out, including less cytotoxic genome editing intermediates and more predictable genome editing outcomes. Finally, base editors can be used to mutate conserved splice acceptor sites within introns, in effect facilitating exon skipping, in a method called CRISPR-SKIP.⁶⁷ This method has the potential to modulate expression of different protein isoforms and skip mutation-containing exons such as in Duchenne muscular dystrophy and Huntingon’s disease.⁶⁸

Base editors provide a potentially less cytotoxic and more efficient method of point mutation introduction than DSB-reliant methods. Nevertheless, this technology faces several limitations as it continues to grow. For example, the current deaminating base editors are only able to facilitate C•G to T•A and A•T to G•C base pair conversions. To be more universally applicable, new base editors must be developed that catalyze additional point mutation changes. Effective intermediates for future base editors will likely be found in less common naturally occurring DNA modifications, as demonstrated by the more predicable editing outcomes observed with base editing using inosine intermediates compared to uracil.¹⁷ Off-target effects are a large concern with genome editing agents in general, and indeed off-target editing is observed with current base editors.⁶⁹ The majority of off-target base editing sites overlap with those of wild-type Cas9. Consequently, base editors derived from Cas9 variants with increased specificity have been shown to alleviate unwanted editing at these sites.⁵⁶ However, off-target sites unique to BE3 have been identified, indicating that creative new solutions are needed in the future to increase base editing specificity. The base editors described here allow for the direct chemical modification of target nucleobases in a programmable manner.

These tools hold tremendous potential for regulating the central dogma. Cytosine and adenosine deamination DNA base editors facilitate the efficient installation of point mutations throughout the genome, allowing researchers to alter the identity or expression levels of proteins and functional RNAs. Cytosine methylation and demethylation DNA base editors allow researchers to modulate gene transcription levels. Adenosine deamination RNA base editors open the door for the targeted modification of RNAs which can ultimately modulate all aspects of transcription and translation. As base editing technologies continue to develop, so will our ability to manipulate the contents of the cell.