Nature Biotech: New tool predicts gene editing success
Since the advent of CRISPR-Cas9 technology in 2012, gene editing has entered the fast lane and achieved a series of new breakthroughs. If CRISPR-Cas9 is compared to molecular scissors capable of destroying target genes, then the base editor (base editor) can be called a molecular pencil because it can replace a single nucleotide. The prime editor developed in 2019 has more powerful functions and can search and replace the genome, which can be called the "word processor" of the molecular world.
The ultimate goal of developing these technologies is to repair harmful mutations in human genes. More than 16,000 small deletion variants are causally linked to human disease and can theoretically be repaired by insertion of the deleted sequence. Cystic fibrosis is a good example, with 70% of cases caused by trinucleotide deletions.
For clinical applications, a technology is needed to insert sequences accurately, efficiently and safely without undesired results. Although the prime editing system has shown great potential in the treatment of genetic diseases such as cystic fibrosis, it is still unclear which factors determine the editing efficiency.
Researchers from the Wellcome Sanger Institute in the United Kingdom and the University of Tartu in Estonia recently published a paper in the journal Nature Biotechnology that they have developed a new tool that can predict the probability of successfully inserting a gene-edited DNA sequence into the genome
Multiple Factors Affecting Insertion Efficiency
In this study, the researchers attempted to systematically assess how the length and composition of the inserted sequence, cell line, target site, and different versions of the prime editor affect insertion efficiency.
To this end, they designed a total of 3,604 pegRNAs encoding insertions upstream of the nicking site, ranging in length from 1 nt to 69 nt and varying in GC content. They inserted sequences into two cell lines (HEK293T and HAP1 cells) targeting four target loci (HEK3, EMX1, FANCF, and CLYBL). A week later, they sequenced the genomes of the cells to see if the edits had been successful. The overall strategy for evaluating insertion efficiency is detailed in the figure below
Because the insertion rate varied across three orders of magnitude, the researchers sought to understand the relevant features, starting with the length of the insertion. They found two features in HEK293T cells: 3 and 4 nt sequences had a higher insertion rate than other sequences; 15-21 nt sequences had a higher insertion rate than surrounding sequences. However, in HAP1 cells, the insertion rate of short sequences of 1-4 nt was not higher than that of longer sequences. They attribute this to the mismatch repair (MMR) system, as HEK293T cells are partially deficient in MMR. This was also demonstrated in HAP1 cells knocked out of the mismatch repair gene MLH1, suggesting that the MMR system hinders the insertion of short sequences.
Afterwards, they analyzed how the different steps of prime editing affect the insertion rate of the sequence. They found that if the pegRNA contained four or more consecutive adenines, the insertion rate dropped significantly. Furthermore, another important step in prime editing is the equilibrium between intermediates with a 5' flap (containing the wild-type sequence) and a 3' flap (containing the insert), while the 5' flap nuclease FEN1 and the 3' flap The nucleases TREX1 and TREX2 mediate this balance. They found that the 3' flap nucleases TREX1 and TREX2 inhibited the insertion of longer sequences.
At the same time, the nucleotide composition and secondary structure of the inserted sequence will also affect the insertion rate. The researchers found that the prime editing system had a clear preference for cytosine. For every 1% increase in cytosine in the inserted sequence, the insertion rate increased by an average of 2.2%. Conversely, the percentages of adenine and thymine decreased the insertion rate at each site. In addition, they found that sequences with higher structural strength were able to be inserted more efficiently.
In doing so, they used an oligonucleotide pool provided by Twist Bioscience. According to Twist, their unique silicon-based DNA synthesis platform can generate more than one million oligonucleotides in a single run, with almost no limit on the number, and the oligonucleotide pool is accurate and uniform, allowing people to be full of confidence in the experimental results. confidence. In addition, multiple vectors and gene fragments used in the experiments were also provided by Twist Bioscience.
Predict the insertion rate of different sequences
Knowing the multiple factors that affect insertion rates, the researchers next wanted to predict the insertion efficiency of different sequences at the same site. They adopted a machine learning method and selected 10 features to train the data, including the length, composition, pegRNA secondary structure, and MMR of the inserted sequence. This method, called MinsePIE, was able to predict the insertion efficiency of the test data well, with a correlation of 0.68
After training on existing data, they tested the MinsePIE model on new data and found that it was able to accurately predict the success rate of multiple insertion sequences. Afterwards, they also experimentally tested the predicted sequences. Codon variants predicted to have higher insertion rates did exhibit higher insertion rates compared to variants predicted to have lower insertion rates, highlighting the advantage of the MinsePIE model for codon optimization. The researchers believe that this computational model can help people choose the most efficient sequence to write into the genome.
Finally, the researchers put forward several suggestions on how to improve the insertion efficiency of the prime editing system. They recommend choosing sequences that are high in cytosine and readily form secondary structures. For pegRNA using the U6 promoter, try to avoid the insertion of adenine. For sequences smaller than 14 nt, temporarily inhibiting MMR or knocking out MLH1 will greatly improve insertion efficiency. Collectively, this work increases our understanding of short-sequence insertion efficiencies, with promise for complex genome engineering and correction of various disease-causing mutations.
