<p>Different versions of the original prime editing system have been developed since its initial introduction. Deciding on what system to use for the application in therapeutic human gene editing, especially concerning the correction of F508del, was the goal of this engineering cycle.</p>
<p>
Since we aim to develop a therapy delivered to the human body, we wanted to obtain high editing efficiency while risking as little off-targets as possible and also reducing the size for improved packability.
</p>
<InfoBoxtitle="Prime Editing"id="prime-editing">
<details>
<summary>
</summary>
<p><b>PE1</b>[1], the first version of the Prime Editor features a Cas9(H840A), a Streptococcus pyogenes Cas9 (SpCas9, hereafter just referred to as Cas9) mutant that only cuts the non-target strand of the DNA template[2], and a wildtype reverse transcriptase from the Moloney Murine Leukaemia Virus (M-MLV RT) connected by a serine and glycine rich flexible linker.</p>
<p><b>PE2</b>[1] improves on this concept by incorporating an improved RT with five mutations improving affinity to the template RNA, enzyme processivity and thermostability (D200N/L603W/T330P/T306K/W313F). This version of the prime editor showed varying improvement of editing efficiency over all tested loci and edits with no apparent downsides. Building on the PE2 system, a smaller version of the M-MLV RT was introduced by Gao et al. (2022)[3]. The RT was truncated by 621 bp through deletion of the RNaseH domain, which originally degrades the RNA template, but is not needed for prime editing. The codon optimized version of this truncated RT prime editor (in literature usually called PE2∆RNaseH) was named <b>PE<sup>CO</sup>-Mini</b> in the paper and will be addressed as such here.</p>
<p>The <b>PE3</b>[1] system, described in the same paper as PE1 and PE2, introduces the use of a second single guide RNA besides the pegRNA which leads to a nick in the strand opposite to the edited strand. This is supposed to improve integration of edits by directing cellular DNA repair systems to use the edited strand as a template for resolving base mismatches. Nicks positioned 3‘ of the edit about 40–90 base pairs from the pegRNA-induced nick were able to further increase editing efficiencies about threefold when compared to PE2, but with a higher range of on-target indels , meaning random Insertions and/or Deletions that appear after faulty repair of double strand breaks in the DNA. PE3b, where the protospacer for the nicking sgRNA lies within the edited regions, decreased the indel rate greatly compared to PE3.</p>
<p><b>PE4</b> and <b>PE5</b>[4] expand the PE2 and PE3 systems, respectively, by co-expressing a dominant negative MLH1 protein (MLH1(Δ754–756), hereafter referred to as MLH1dn). The MLH1 protein plays a crucial role in the mismatch repair (MMR) mechanism of the human cell[5] by recruiting other repair proteins and facilitating catalytic function. The mutant still recruits other factors but is impaired in its endonuclease function, disrupting function of the entire repair mechanism. This leads to an average 7.7-fold and 2.0-fold increase in editing efficiency, respectively, compared to PE2 and PE3. This is possibly due to slower repair of mismatches and thus more time for the proteins encoded by LIG1 and FEN1 genes to excise the non-edited 5’ flap and ligate the nick in the edited strand. Additionally, MLH1dn co-expression slightly reduced on-target indels as well as unintended editing outcomes in PE3 systems and did not lead to higher off-target indel rates or overall mutation rates.</p>
<p>With <b>PEmax</b>[4], the structure of PE2 is further enhanced by using human codon-optimized RT, a new linker containing a bipartite SV40 nuclear localization sequence (NLS)[[6]], an additional C-terminal c-Myc NLS[7] and R221K N394K mutations in SpCas9 previously shown to improve Cas9 nuclease activity[8]. These changes led to moderate improvements in editing efficiency compared to previous editor architectures.</p>
<p><b>PE6</b>[9] was made by improving the reverse transcriptase domain of the prime editor using Phage-Assisted Continuous Evolution (PACE). Multiple RT mutants (PE6a-d), derived from RTs of Escherichia coli Ec48 retron, Schizosaccharomyces pombe Tf1 retrotransposon and Moloney Murine Leukaemia Virus, were identified to increase editing efficiency over and/or were smaller than the M-MLV RT used in previous PE systems. Especially <b>PE6c</b> (evolved Tf1 RT) and <b>PE6d</b> (evolved M-MLV RT) showed significantly higher editing efficiencies than even PEmax depending on the targeted loci, with PE6d showing benefits especially in loci forming more complex secondary structures. Recent advancements in prime editing targeting the CFTR F508Δ mutation showed that PE6c was the most promising for editing in this loci[10]. Improvements of nCas9 on the other hand (PE6e-g) were only marginal and highly site specific. All PE6 systems use nicking gRNAs (PE3) by default, but do not co-express MLH1dn.</p>
<p><b>PE7</b>[11] adds an additional RNA binding domain to the Prime Editor. The domain is derived from the La Protein (La(1-194)), an endogenous eukaryotic protein involved RNA metabolism and known for its role in binding polyuridine (polyU) tails at the 3’ ends of nascent transcripts, thus protecting them from exonuclease activity. PE7 showed considerable improvements over PEmax at different loci and different types of edits when used with the PE2 strategy (no nicking gRNAs, no MLH1dn co-expression). Notably, PE7 did perform worse when used with engineered pegRNAs than with regular ones (see pegRNA design).</p>
</details>
</InfoBox>
</div>
<divclassName="box">
<pid="pe1">
<H3text=""id="text"/>
<H3text="PE2 and PECO-Mini"id="text"/>
<H4text="Design"id="text"/>
<p></p>
<p>
For our initial approach, we wanted to start from the beginning and use the PE2 prime editing system. Since our goal of stripping the size of the prime editor was a big factor from the beginning, we did a researched into that direction and found a truncated version of M-MLV RT, PE<sup>CO</sup>-Mini. We then ordered the plasmids for both PE2 and PE<sup>CO</sup>-Mini. Since the PE<sup>CO</sup>-Mini plasmid had a different promotor than pCMV-PE2, we decided to clone the PE<sup>CO</sup>-Mini RT into the pCMV-PE2 vector to allow for direct comparison.
</p>
<H4text="Build"id="text"/>
<p></p>
<p>
We designed primers for the amplification of PE<sup>CO</sup>-Mini RT and cloned it into pCMV-PE2 via double digestion and Gibson assembly.
</p>
<H4text="Test"id="text"/>
<p></p>
<p>
To compare the prime editing performances of M-MLV RT (PE2) and PE<sup>CO</sup>-Mini RT, both were tested using a 2in1 prime editing reporter plasmid system[12] (see Prove of Concept[link]) in HEK293 cells. Contrary to the findings of Gao et al., here the PE<sup>CO</sup>-Mini prime editor performed a lot worse than the PE2 prime editor.
</p>
<H4text="Learn"id="text"/>
<p></p>
<p>
Since we knew, that for a successful therapy targeting the F508del mutation a very high prime editing efficiency was of utmost importance, we decided against using PE<sup>CO</sup>-Mini as the basis for our approach and that we have to look for other alternatives.
</p>
</p>
</div>
<divclassName="box">
<pid="pe2">
<H3text=""id="text"/>
<H3text="PE6c"id="text"/>
<H4text="Design"id="text"/>
<p></p>
<p>
During our initial talk with Mattijs Bulcaen[Link], he recommended a talk of David Liu[Link] at an online conference, where he presented unpublished data about his laboratory working on prime editing for F508del correction. We investigated it and through this came across the PE6 generation of prime editors. Seeing that the Liu Laboratory eventually decided on using the PE6c system, we adopted the findings.
</p>
<H4text="Build"id="text"/>
<p></p>
<p>
We got the plasmid carrying the PE6c prime editor. Except for the RT and a few improving mutations in the Cas9 enzyme, it has the same architecture as PE2, which made comparison quite easy.
</p>
<H4text="Test"id="text"/>
<p></p>
<p>
We tested PE6c against PE2 using the same reporter system as mentioned above for PE<sup>CO</sup>-Mini. PE6c, as expected from the literature, proved way more efficient in prime editing.
</p>
<H4text="Learn"id="text"/>
<p></p>
<p>
The data from literature as well as our own experiments confirmed that PE6c architecture is superior to PE2 even without using nicking gRNAs that help suppress mismatch repair. This led us to the decision to use the PE6c reverse transcriptase and parts of the overall architecture for our subsequent tests.
</p>
</p>
</div>
<divclassName="box">
<pid="pe3">
<H3text=""id="text"/>
<H3text="PreCyse Casette"id="text"/>
<H4text="Design"id="text"/>
<p></p>
<p>
In the later stages of our project, the Liu laboratory published their own findings regarding CFTR F508del targeting with prime editing[10]. The data showed that the editing efficiency of PE2 based systems, even when using PE6c reverse transcriptase, might not be sufficient for application in a therapy. Also, the plasmids of current prime editors did not include restriction sites that would have allowed replacing components like the nickase to test alternatives. This is why, in a cherry-picking manner, we combined the PE6c architecture prime editor with the most promising aspects of other prime editors, creating the PreCyse cassette.
</p>
<p>
Our decision on what components of existing prime editors we wanted to use was mainly driven by two factors: efficiency and precision. In prime editing, these two are often opposing forces, which means advancements improving efficiency often also increase the risk of off-targets mutations and on-target undesired editing. For this reason, we decided against using nicking gRNAs. Although they have been proven to reliably improve editing efficiency, they increase the risk and possible scope of off-target cleavage and mutations. Additionally, if <b>PE3b</b> is not applicable, there is a chance for double strand breaks to occur, which diminishes the safety advantage of prime editing over other common CRISPR-based methods. Co-expression of MLH1dn can improve editing efficiency in the same way as nicking gRNAs do, by helping to evade of the cellular mismatch repair mechanisms. The use of MLH1dn is especially impactful, when nicking gRNAs are not used, which is perfect in our case. Recently, the La poly(U)-binding motif has been shown to enhance prime editing efficiency, presumably through protection of the 3’ poly(U) tail of the pegRNA from RNases. The motif is also comparatively small, which aligns with the overall goal to create a compact prime editing tool. This is why PreCyse Casettes have been designed to include the La RNA binding motif fusion and the dominant negative MLH1 protein.
</p>
<H4text="Build"id="text"/>
<p></p>
<H4text="Test"id="text"/>
<p>
The PreCyse cassette comes in three versions: PreCyseA, the most basic version, comprises of a T7 promoter and an open reading frame, which includes NLS and one typeIIS restriction enzyme cloning site for a nickase and a reverse transcriptase each. For possible future additions like e. g. selection markers, a BamHI restriction site at the end of the coding sequence allows for easy in-frame Gibson cloning. Building on this basis, PreCyseB expands PreCyseA by the La Poly(U)-binding motif. PreCyseC additionally introduces the co-expressed MLH1dn. The cassettes were ordered in three individual parts to be put together with a pCMV-PE6c backbone via Gibson Cloning in different configurations to create the three variants. In the plasmid the cassette is expressed under a CMV promoter and followed by a polyadenylation signal. The PreCyse Casettes themselves can be used as a BioBrick RFC[10] standard compatible composite part can thus be freely combined with other parts. The nickase and RT slots can be used for inserting any basic or composite part compatible with the Type IIS RCF[1000] standard for fusion proteins. The PreCyse Casette is meant to be a contribution to the iGEM community and a base for other teams to join us and researchers around the world to innovate in the exciting field of prime editing.
