Engineering structural variants with prime editing

CRISPR/Cas-based gene editing has revolutionized biological research by enabling programmable and scalable manipulation of genomic sequences. Prime editing has emerged as a powerful tool to introduce virtually any type of edit with high precision, but editing efficiency varies between edit types. While first insights into determinants have been gained for small edits, dual-guide approaches enabling larger edits remain unexplored—posing a challenge for large-scale applications and functional screens. In this thesis, I explore features determining prime editing efficiencies for both insertions and long deletions. After demonstrating that paired prime editing screens work efficiently across reporter locations, I scaled paired prime editing to regions across the genome, thereby identifying essential coding and non-coding sequences. First, I built a predictive model for prime editing insertion efficiencies using experimental data generated in our laboratory. This model, called MinsePIE, forecasts insertion rates of sequences across various lengths and nucleotide compositions. By characterizing the importance of features on the model output, I identified key determinants of editing efficiency, such as length and structure of the reverse transcriptase template. I validated the model across internal and external datasets on novel target sites and insertion sequences and finally demonstrated its application for optimizing prime editing experiments by predicting optimal insert sequences for protein tags and padding sequences. While prime editing has been well optimized for single-guide edits, the principles for generating large deletions have remained uncharacterized at scale. To address this gap, I systematically measured the rate of 3,881 deletions with paired prime editing spanning up to 1.2 Mb, identifying factors like deletion length, target site contact frequency, and individual pegRNA efficiency as key determinants of editing efficiency. While the frequency of deletions decreases with length, megabase-scale deletions can still be achieved at single-digit rates. This work provides foundational guidelines for generating long deletions with paired prime editing. Finally, I applied paired prime editing to systematically interrogate the essentiality of non-coding DNA by generating a total of 11,084 tiling deletions in regions surrounding 149 genes. While many of the non-coding regions were dispensable, the screen revealed a subset of deletions with a significant impact on cell survival. This highlights the potential of prime editing for studying non- coding DNA, regulatory elements, and genome architecture at scale. In my PhD, I established and characterized a toolbox to perform precise insertions and deletions using prime editing. I envision that these technologies will facilitate the functional screening of genomic sequences beyond coding genes and the systematic exploration of structural variants.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2025

Author dc:creator

• Weller, Juliane

Advisor dc:contributor.advisor

• Parts, Leopold

• genome engineering

Licence

• https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/cb2d432e-2821-4ac5-b339-e8c080c925c2/download
• http://purl.org/NET/rdflicense/allrightsreserved

DOI dc:identifier.doi

https://doi.org/10.17863/CAM.118838

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/385050