(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



In the business of base editors: Evolution from bench to bedside [1]

['Elizabeth M. Porto', 'Department Of Chemistry', 'Biochemistry', 'University Of California', 'San Diego', 'La Jolla', 'California', 'United States Of America', 'Alexis C. Komor']

Date: 2023-04

Top left : CBE architecture shown with principal components: Cas9n in grey (outline of crystal structure obtained from PDB: 6VPC), CBE deaminase APOBEC3A in red (outline of crystal structure obtained from PDB: 5SWW), and UGI in purple (outline of crystal structure obtained from PDB: 1UGI). The deaminase and UGI components are tethered to nCas9 via short amino acid linkers (grey). Overlayed on top of the principal components is a general schematic of the mechanism of action; the gRNA (brown) will bind to the DNA protospacer (sequence of 20 nucleotides proximally located to the 3-nucleotide PAM (violet) sequence), in the process exposing a single-stranded DNA “bubble” open for cytosine deamination. Deamination produces a U•G intermediate, which is processed by the cell to produce an overall C•G to T•A conversion (shown in base conversion inset). Concurrently, nCas9 will nick the unedited DNA strand (blue triangle) to increase editing efficiency. Similarly, the addition of the UGI component increases editing efficiency. Top right : ABE architecture, simplified mechanism schematic, and overall base conversion are shown. Key differences of the ABE architecture are as follows: ABE deaminase TadA-8e, similarly in red, (outline of crystal structure obtained from PDB: 6VPC) replaces CBE deaminase and the lack of a UGI component, as ABE utilizes an inosine intermediate, compared to the CBE architecture. Bottom : A noncomprehensive sampling of notable variations on key CBE and ABE principal components are shown. Collectively, these substitute components serve an array of purposes including increased on-target editing, decreased off-target editing, and relaxed PAM requirements for broadened utility. Development of new and enhanced base editor principal components is a populated field of study with new results being published rapidly. ABE, adenine base editor; AID, activation-induced cytidine deaminase; AmAPOBEC1, Alligator mississippiensis APOBEC1; APOBEC, Apolipoprotein B mRNA editing enzyme, catalytic polypeptide; CDA1, cytidine deaminase 1 sourced from sea lamprey; Cas9n, Cas9 nickase; CBE, cytosine base editor; dxCas9, catalytically dead expanded PAM Cas9; eCas9, enhanced specificity Cas9; enAsdCas12a, enhanced Acidaminococcus sp. BV3L6 catalytically dead Cas12a gRNA, guide RNA; hAPOBEC, human APOBEC; HiFi Cas9, high fidelity Cas9 variant; HF-Cas9, high fidelity Cas9 variant; HypaCas9, hyper accurate Cas9; LbdCas12a, catalytically dead Lachnospiraceae bacterium Cas12a; NLS, nuclear localization signal PAM, protospacer adjacent motif; PpAPOBEC1, Pongo pygmaeus APOBEC1; rAPOBEC1, Rattus norvegicus APOBEC1; RrA3F, Rhinopithecus roxellana APOBEC3F; SaCas9, Staphylococcus aures Cas9; SECURE, selective curbing of unwanted RNA editing; SpCas9, Streptococcus pyogenes Cas9; ssAPOBEC3B, Sus scrofa APOBEC3; UGI, uracil glycosylase inhibitor.

To address this, a second-generation CBE was developed, BE2, which incorporated a uracil glycosylase inhibitor (UGI) peptide to temporarily block BER, thus preventing uracil excision and increasing C•G to T•A conversion efficiencies. One last modification to the system was to exchange dCas9 for a nickase version of the enzyme (nCas9) and produced the final original CBE, named BE3. BE3 installs a DNA nick on the strand opposite the uracil-containing strand. This in turn manipulates the cell’s native DNA repair processes to preferentially replace this strand and use the uracil-containing strand as a template, thus increasing editing efficiency even more ( Fig 1 ). Shortly after the development of BE3, an additional CBE (named Target-AID) was described, which included similar components (a cytidine deaminase, nCas9, and UGI), but utilized the more active cytidine deaminase pmCDA1 (cytidine deaminase 1 sourced from sea lamprey) and fused together in a different orientation, resulting in a slightly shifted editing window compared to BE3 [ 5 ]. Target-AID demonstrated the robustness of this general strategy for targeted, programmable point mutation introduction.

Following formation of the Cas9:gRNA:DNA ternary complex, a subset of one DNA strand is now single-stranded and accessible to rAPOBEC1 for deamination chemistry ( Fig 1 ). Cytidines that are within this “editing window” are deaminated by rAPOBEC1, which produces a C•G to U•G conversion. The development and characterization of many subsequent CBEs have revealed that several factors influence which nucleotides within the protospacer comprise this “editing window”, and include the Cas homolog that is used, the linker length and composition between the deaminase and Cas protein, the overall architecture of the base editor, and the deaminase enzyme used (discussed later). For BE1, the deamination activity window is between positions 4 to 8 within the protospacer ( Fig 1 ). Processing of the U•G intermediate by the cell, using the U-containing strand as a template, results in an overall C•G to T•A conversion. However, the presence of the U•G mismatch intermediate triggers the cell’s native base excision repair (BER) pathway to excise the uracil and revert the intermediate back to the original C•G base pair [ 4 ]. Consequently, editing activity by BE1 in live mammalian cells was quite low, and C•G to non-T•A conversions were observed as well (discussed later).

Currently, two classes of base editors exist: cytosine base editors (CBEs) and adenine base editors (ABEs). In the first example of targeted point mutation introduction via a non-DSB mechanism, the original CBE (named BE1) was created by fusing a catalytically inactive or “dead” Cas9 (dCas9) enzyme with the naturally occurring cytidine deaminase enzyme APOBEC1 (rAPOBEC1 sourced from Rattus norvegicus) [ 1 ]. The dCas9 protein complexes with a preprogrammed guide RNA (gRNA) and subsequently locates and binds to a specific DNA sequence (the protospacer) through the formation of an R-loop, driven by base-pairing between the protospacer and the first 20 nucleotides of the gRNA (the spacer; Fig 1 ) [ 2 ]. For the gRNA to bind to the protospacer, the protospacer must also be immediately adjacent to a protospacer adjacent motif (PAM) sequence ( Fig 1 ). In the case of the widely used Streptococcus pyogenes Cas9 (spCas9), the PAM sequence is 5′-NGG-3′, which has been calculated to occur once every approximately 42 bases throughout the human genome [ 3 ].

Directed evolution facilitates the enhancement or alteration of the activity of a given protein [ 9 – 11 ]. The protein of interest is mutagenized to produce a library of members, and active members are screened or selected to identify those with the new or enhanced activity of interest. To generate the first ABE, TadA, a tRNA adenosine deaminase sourced from Escherichia coli, which shares partial structural homology with the rAPOBEC1 enzyme employed by CBEs, was selected as a starting point. Over the course of seven rounds of directed evolution, ecTadA accumulated fourteen mutations to produce ABE7.10, which demonstrated on average 58% A•T to G•C editing efficiency across a variety of target sites with various sequence contexts [ 8 ]. It is important to note that adenine base editing did not require any BER inhibition components (such as the UGI of the CBE), presumably due to a lower efficiency of inosine excision by BER glycosylase enzymes. Consequently, no A•T to non-G•C editing was observed by ABE7.10.

Using CBEs as a model, researchers sought to expand the base editor toolbox to include ABEs, which would use adenosine deamination chemistry to install A•T to G•C base pair conversions using an inosine-containing intermediate. ABEs would be capable of correcting the most common pathogenic single nucleotide variant (SNV), making them a vital tool for therapeutic genome editing [ 6 , 7 ]. While the general approach of replacing rAPOBEC1 for an ssDNA-specific adenosine deaminase enzyme was simple and elegant, unfortunately, no such naturally occurring enzyme existed, and it therefore needed to first be created.

Limitations and modifications

We focus here on the limitations of base editing tools from a therapeutic perspective and the corresponding modifications to the original ABE and CBE constructs that have been engineered to overcome these limitations. The most obvious and major restriction of base editing technologies is the limited types of base pair conversions (C•G to T•A and A•T to G•C only) achievable with CBEs and ABEs. Expansion of the base editor toolbox in this area has been via the development of “glycosylase base editors,” which utilize the basic CBE architecture with additional enzyme components that facilitate excision of the uracil intermediate. Specifically, a suite of “CGBEs” (C•G to G•C base editors) has been developed, which exclude the UGI component of the CBE architecture and instead incorporate a uracil glycosylase enzyme and/or error-prone polymerases [12–16]. In these editors, the uracil intermediate is efficiently excised by either the endogenous uracil glycosylase enzyme of the cell, or that included in the CGBE architecture, to produce an abasic site. The resulting abasic site is then processed by the translesion synthesis pathway of the cell, or the polymerase included in the CGBE architecture, to mutagenize the target base, with a C•G to G•C base pair as the most common overall outcome. One such glycosylase base editor is currently being used in a clinical trial by Bioray Laboratories (discussed below). This same strategy was recently applied to ABEs as well, where an engineered hypoxanthine glycosylase enzyme (derived from N-methylpurine DNA glycosylase, MPG) was fused to an ABE, resulting in an adenine transversion base editor (AYBE) that mutagenizes target adenines, with an A•T to C•G base pair as the most common overall outcome [17].

An additional major limitation of early base editors was their targeting scope. Due to the restrictive editing window (positions 4 through 8 in the most widely used editors), many times a requisite PAM sequence could not be located at the necessary location. After establishing the architectural framework of the first CBE, subsequent efforts found that replacing the Cas9 enzyme with Cas9 variants with relaxed or altered PAM requirements, or Cas homologs from different species, resulted in editors with high editing efficiencies and significantly increased the targeting scope [18,19]. With the advent of extremely PAM-relaxed Cas9 variants, such as Cas9-NG and SpRY-Cas9, base editor targeting scope issues have been largely alleviated [20,21]. ABE7.10 was not as compatible with alternative Cas proteins, but this issue was resolved with the development of next-generation ABEs (discussed next).

An important characteristic of a therapeutic genome editor is high editing efficiency. Additional directed evolution efforts have been undertaken on both CBEs and ABEs to improve their overall efficiencies and remove sequence context biases that the deaminases possessed. Architectural engineering efforts on the original BE3 construct produced BE4, which has higher editing efficiencies and product purities than BE3 [22]. In fact, BE4 is currently being used in a clinical trial by Great Ormond Street Hospital for Children (GOSH; discussed below). Directed evolution efforts have also produced optimized CBEs via the improvement of deaminase kinetics and/or solubility [23,24]. Additionally, codon optimization is crucial for optimizing expression of BEs in different cell types, which is an important consideration therapeutically [25,26]. The further directed evolution of ABE7.10, resulting in various ABE8 and ABE9 constructs, was particularly important from a therapeutic context, as the resulting ABE8 variants are being used in the current clinical trials [25,27,28]. As mentioned previously, these ABE8 variants are also compatible with additional Cas homologs, which in effect expanded the targeting scope of these editors significantly. In fact, ABE8 variants are currently being used in two clinical trials (discussed below).

Finally, arguably the most important limitation of base editors from a therapeutic perspective are unintended edits. Unintended edits include any modification to the cell’s genome other than the intended edit. These may include “bystander edits” (which occur within the same protospacer as the intended edit), the wrong type of edit being installed at the target nucleotide (such as C•G to non-T•A conversions by CBEs) or “off-target edits” (which occur at other genomic loci in the cell), and it is important to note that these unintended editing events aren’t necessarily deleterious, and in fact many times can be benign. Bystander editing occurs as a consequence of deaminase processivity; if multiple target Cs or As are accessible within the ssDNA window, the deaminase will modify some or all. However, extensive deaminase engineering efforts have resulted in less-active deaminases that have narrower activity windows. Additionally, alteration of the overall architecture can manipulate the activity window. Furthermore, with the development of PAM-relaxed Cas9 variants, multiple gRNAs can be designed for a given target base, some of which will “push” the bystander bases outside of the editing window. These efforts are more thoroughly outlined in several key publications [22,25,29–31]. An additional type of bystander editing was observed with ABEs, namely, cytidine deamination activity by the mutant TadA protein, which would result in undesired bystander C•G to T•A mutations in addition to the desired A•T to G•C mutation [32]. This activity was then significantly reduced through engineering efforts, resulting in more precise ABE variants [33].

Extensive work has been done to characterize the off-target editing efficiencies of base editors, and three different types have been observed: gRNA-dependent off-targets; gRNA-independent DNA off-targets; and gRNA-independent RNA off-targets [34–38]. gRNA-dependent off-target editing occurs when Cas9 binds to a homologous genomic locus despite mismatches between the protospacer and spacer. The use of “high fidelity” Cas variants, which have lower tolerance for mismatches, can be incorporated into the base editor architecture to eliminate these [39]. Additionally, judicious choice of the gRNA can sometimes eliminate potential off-targets. gRNA-independent off-target editing occurs when the deaminase has access to ssRNA (both ABE and CBE) or ssDNA (CBEs only) within the cell (such as mRNA and transcription or replication bubbles) and deaminates cytosines or adenines within these bubbles. Several key publications have reported engineering of the deaminase domain (in both ABEs and CBEs) to reduce or eliminate RNA off-target editing events [40–43]. To reduce DNA off-target editing events, researchers have mutated the rAPOBEC1 protein to reduce its catalytic activity, as well as identified APOBEC homologs that naturally have lower gRNA-independent off-target editing activities [44,45]. Additionally, researchers have leveraged the previously undesired cytidine deamination activity observed with ABEs to engineer TadA-derived CBEs that have no gRNA-independent off-target DNA editing activity, like their ABE counterparts [46–48].

Notably, delivery of base editors as mRNA rather than plasmid DNA significantly reduces all forms of off-target editing [28]. A relationship between genome editing specificity and delivery modality/dosage was discovered prior to the development of base editors [49–52]. Genome editing agents typically modify the on-target locus first and will then modify off-target loci if their intracellular lifetime is long enough. To balance high on-target editing with minimal off-target editing, a short burst of a high level of active editor complex is therefore desired. Delivering DNA encoding for the editor will result in long-term expression, increasing chances of off-target editing. The lifetime of RNA is shorter than that of DNA, and transcription is not required to produce active editor when delivering mRNA encoding for the editor and gRNA. This results in a shorter timeframe between delivery and editing for mRNA and gRNA versus DNA, as well as shorter-term expression of active editor. Both mRNA and gRNA can be chemically modified to extend their half-lives as well. In a recent example of ex vivo base editing in hematopoietic stem cells (HSCs), chemically modified mRNA encoding BE3 and synthetic gRNA were electroporated, and BE3 protein expression peaked at 12-hour post-electroporation and was nearly entirely gone by 24-hour post-electroporation [53]. Furthermore, delivery of genome editing agents as purified protein:gRNA complexes (discussed below) results in the shortest overall lifetime of active editing agents. However, large-scale production of base editors at the purity required for therapeutic applications has been challenging, and thus mRNA delivery of base editors is generally preferred [54]. While off-target DNA edits are a therapeutic concern (particularly if they happen to occur in oncogenes or tumor suppressor genes), the quick turnover of mRNA within the cell alleviates some concerns regarding RNA off-targets.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002071

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.