Engineering Success

Once the concept of molecular recording using sequential edits on an RNA tape was formed, we needed to choose the most suitable editing system. Through extensive research on various RNA-editing systems (refer to our Human Practices page for more details on our research sessions and expert consultations), we ruled out the LEAPER system, which leverages endogenous ADAR for programmable editing. LEAPER uses 100-200 nucleotide long gRNAs, which were too lengthy for our sequential recording tape design. We also considered RESCUE (RNA Editing for Specific C to U Exchange) (Abudayyeh et al., 2019), but its mechanism was incompatible with our readout system, which relies on the destruction of START codons. After several cycles of research sessions, we chose the REPAIR (RNA Editing for Programmable A to I Replacement) system (Cox et al., 2017), as it supported the sequential destruction of START codons without requiring large aptamer sequences. Since REPAIR v2 has been shown to be more efficient than the first version (Cox et al., 2017), we decided to begin our experiments with it.

To ensure the most efficient rate of RNA editing from the REPAIR v2 system, before proceeding with our planned tape-switching experiments, we decided to optimize the ratio of transfected plasmids ratios in our assays.

After choosing a base-editing system, we needed to select the most suitable backbone to insert our recording tape into. To minimize potential cross-interactions with our sensitive recording tape, we decided to design and clone our own minimal vector, which we refer to as pRAM (BBa_K5102000).

In addition to our RNA-editor and backbone-plasmid, our main goal was to design a synthetic recording tape that would feature a series of START codons in three different open reading frames. This design would allow for the downstream placement of fluorescent proteins, serving as an in vivo read-out system for the cellular events we intend to record. Since we had to design the recording tape from scratch, we started with a minimal design, which included only Kozak sequences and START codons placed in these three reading frames. In between are stretches of nucleotides that can be replaced by any of the three nucleotides: G, C, or U. The dPspCas13b/ADAR2DD fusion is directed to the target site by a guide RNA (gRNA). The gRNA is composed of spacers and a hairpin sequence essential for dPspCas13b function. So as a next step, we had to design the binding region of the spacer gRNA sequences to be complementary to the target with two mismatches: one that would disrupt unspecific binding, and an A to C mismatch at the target deamination site (See our parts page for more information on the tape-design process). The hairpin sequence was identical to the one used in the REPAIR v2 system developed by Cox et al., 2017.

After realizing the limitations of manual nucleotide replacement, we decided to write specific code for designing the recording tape and guide RNAs (gRNAs) that would enable us to include more functions and parameters that we could manually optimize for. Apart from correct folding of gRNAs and its specificity, we included functions addressing tape accessibility (by preventing the formation of unwanted secondary structures within the tape sequence), selectivity (by enhancing the binding affinity of gRNAs to their targets within the tape while minimizing affinity for sequence-similar mismatches), precision (by control nucleotide composition to minimize unintended local deamination events), and compatibility with BioBrick standards.

After obtaining sequences for our recording tapes, we proceeded to evaluate the switching states from state zero to one. Here, state zero represents a condition where none of the adenosines in START codons are deaminated, allowing translation of proteins in the first reading frame. State one occurs after the deamination of the first adenosine in the tape and is characterized by the disruption of the first START codon. This leads to a switch of the open reading frame by one. To preliminarily validate these switches in state, we designed a NanoLuciferase assay where the NanoLuc protein sequence was positioned in the second open reading frame.

Following preliminary experiments with NanoLuciferase, we proceeded to validate our primary readout system involving fluorescent proteins (XFPs). We first designed composite parts tailored to visualize tape state switching effectively, visualized via a switch in fluorescence protein expression, with eUnaG (a fluorescent protein that can be translated in all reading frames) as a protein expression control. Additionally, in this experimental setup we extended our design to test multiple state switching events, with the addition of all gRNAs guiding the dPspCas13b/ADAR2DD system to perform three sequential edits.

Each recorded state of the tape is visualized by expression of one of three fluorescent proteins. We selected three fluorescent proteins — miRFP670nano3, mScarlet-3, and mTagBFP2 — based on criteria such as emission and excitation wavelengths, maturation time, and fluorescence intensity. We opted to choose three fluorescent proteins that have distinct excitation and emission wavelengths, so that the three recording states can be exactly differentiated. Furthermore, we also had to avoid the excitation and emition specter of eUnaG, a fluorescent protein expressed in every state for RNA-trancription referance. We also opted for proteins with a similar fluorescence intensity to have comparable measurements of the states. However, for preliminary analysis, we decided to test imaging of XFPs with similar excitation/emission wavelengths that were available to us in the lab: miRFP670nano3, mScarlet-I, mTagBFP2.

After selecting the XFPs for our project, we performed in silico design of our reporter composite part, which will ultimately be positioned downstream of our designed recording tapes.

The decision on which amino acids to mutate in order to address STOP codons in one of the forward frames was based on amino acids structure similarities, literature research and alignment of proteins within the same family. Lysine in the 16th position of mTagBFP2 has been mutated to arginine based on biochemical similarity of those amino acids (K16R). M22V mutation in mScarlet-3 has been performed based on homology to mRed7, whereas N98V based on structural similarity of amino acids.

We then need to codon optimize our proteins in all three forward open reading frames for expression in HEK293T cells. While codon optimization tools are easily accessible, none of them allow for optimization in more than one frame. Therefore, we needed to create our own tool tailored to the project. This task required extensive research on the mechanisms of translation efficiency, codon usage bias, and mRNA stability specific to HEK293T cells. We had to ensure that the tool could optimize codon usage across all three reading frames without altering the amino acid sequence, while also eliminating problematic motifs such as premature stop codons, splice sites, and restriction sites. Additionally, controlling factors like GC content and preventing secondary structures that could trigger mRNA degradation pathways were critical to the success of the project.

During one of our iHP (Integrated Human Practices) meetings (learn more about the expert meetings on our HP wiki page), we realized that while our code effectively excludes canonical splice sites, non-canonical splice sites are more challenging to screen for.

The next stage of our engineering cycle was to design cloning of fluorescent proteins into our minimal synthetic vector, pRAM. Each composite included a T2A site, one of the three XFPs of miRFP670nano3, mScarlet-3, mTagBFP2, P2A site and eUnaG.

After successful imaging of single XFPs, we proceeded with the design of the final composite parts that include the recording tape-T2A-miRFP670nano3-P2A-eUnaG-T2A-mScarlet3-P2A-eUnaG-T2A-mTagBFP2-P2A-eUnaG.

After successful cloning, we moved on to validate the functionality of our readout system, which is dependent on the recorded state of the tape. We tested the ability to switch between one or more states by introducing all gRNAs designed to guide dPspCas13b/ADAR2DD for sequential RNA edits.

While testing our readout system using tapes in state zero was crucial, we wanted to go one step further in testing the complete recording system. We identified the need for additional controls with tapes that had already been modified once (state 1) or twice (state 2).

During the development of our project, we explored several strategies to improve the stability of our RNA recording tape. These concerns were brought into discussion during several of our research sessions, as well as mentioned in our meetings with experts, and raised in specific and more general (feasibility) terms on our survey on molecular recording. More details on all these topics can be found in our STIR protocols and overall Human Practices page.

In addition to increasing the stability of our RNA construct, we aimed to enhance its expression. During an iHP meeting with Prof. Danny Nedialkova, it was suggested that incorporating a strong 5’ UTR could help achieve this. While exploring the idea, we came across a study demonstrating the use of synthetic 5’ UTRs to improve mRNA translation for gene editing delivered via mRNA.

After verifying the synthetic UTRs, we selected the highest-scoring UTR for cloning into our construct, which includes the recording tape and reporter system with XFPs. Cloning primers were designed to ensure that the synthetic UTR is placed on primer overhangs, allowing the primers to overlap and facilitate a Gibson assembly.

After integrating the necessary elements to enhance the expression and stability of our RNA construct in the plasmid, we aimed to test its performance within a cellular environment. To achieve this, we developed a predictive mathematical model to simulate RNA transcription and degradation dynamics in living cells. This model allows us to understand how our construct behaves over time and to optimize its performance accordingly.