Precision multiplexed base editing in human cells using Cas12a-derived base editors

Abstract

Base editors enable the direct conversion of target nucleotides without introducing DNA double strand breaks, making them a powerful tool for creating point mutations in a human genome. However, current Cas9-derived base editing technologies have limited ability to simultaneously edit multiple loci with base-pair level precision, hindering the generation of polygenic phenotypes. Here, we test the ability of six Cas12a-derived base editing systems to process multiple gRNAs from a single transcript. We identify base editor variants capable of multiplexed base editing and improve the design of the respective gRNA array expression cassette, enabling multiplexed editing of 15 target sites in multiple human cell lines, increasing state-of-the-art in multiplexing by three-fold in the field of mammalian genome engineering. To reduce bystander mutations, we also develop a Cas12a gRNA engineering approach that directs editing outcomes towards a single base-pair conversion. We combine these advances to demonstrate that both strategies can be combined to drive multiplex base editing with greater precision and reduced bystander mutation rates. Overcoming these key obstacles of mammalian genome engineering technologies will be critical for their use in studying single nucleotide variant-associated diseases and engineering synthetic mammalian genomes.

Fig. 1. Engineering Cas12a for multiplexed base editing.

A The ability of Cas12a to process gRNA arrays without accessory factors allows for the efficient delivery of many gRNAs into a single cell. Truncated gRNAs can be used to prevent deamination of neighboring bases. Created in BioRender. Schweitzer, A. (2025) https://BioRender.com/z2a6bsw. B Comparison of two published dLbCas12a-derived CBE and (C), one published dLbCas12a-derived ABE system for MBE in HEK293 cells. Heatmaps show normalized mean %T/%G values from three independent replicates (n = 3). Normalization was performed by subtracting the mean %T/%G values of the nt-ctrl condition from the mean %T/%G values of the experimental condition. Only positions 7–12 for CBEs and positions 8–12 of ABEs are shown, as those correspond to the editing window of the used systems. sg: single guide, dg: double guide, tg: triple guide. Source data are provided as a Source Data file.

Fig. 2. BEACON2 mediates highly multiplexed base editing.

A Schematic of the architecture of the three gRNA expression plasmids used in this study. SV40: Simian Virus 40 termination signal. B Schematic of gRNA array architectures used in this study. The standard array consists of the DR and guide sequences and does not include any additional sequence elements. The SynSep array includes an AAAT sequence upstream of each DR sequence, and the VarSep arrays include variable 4 nt sequences upstream of each DR sequence (see Supplementary Data 1). C Comparison of different array architectures for BEACON2-mediated multiplex editing of 14 RUNX1 target sites. D Editing frequencies of gRNAs across arrays shown in (C), normalized to their efficiency when expressed using the standard array architecture. E BEACON2-mediated multiplex editing of 3–16 target sites, across six genes located on five chromosomes. F, Comparison of a Pol-III promoter (hU6) and two Pol-II promoters (CMV and EFs1a) for the expression of the standard gRNA array shown in (D). All data are the mean ± SD of three independent replicates (n = 3). C–F show normalized mean %T values at the highest edited C for each gRNA as determined based on the single gRNA condition. Normalization was performed by subtracting the mean %T of the non-targeting (nt-ctrl) condition from the mean %T of the experimental condition. Source data are provided as a Source Data file.

Fig. 3. Genome wide off-targets and bystander edits generated in BEACON2 treated cells.

A Total number of mutations found in three edited clones and one clone transfected with a non-targeting (nt-ctrl) gRNA. B Total number of detected single-nucleotide variants (SNVs) by base-change for three edited clones and one clone transfected with an nt-ctrl gRNA. C Summary of allele frequencies observed at the 16 targeted sites after editing with the 16x gRNA array (Fig. 2E), showing combined outcomes based on the number of edited Cs. Data represents three cell clones isolated from the edited population. D Total number of edited target sites per clone. E Overlap of C > T/G > A SNVs detected in our analysis and predicted off-target sites by Cas-OFFinder.

Fig. 4. Engineered gRNAs reduce frequencies of bystander mutations mediated by BEACON2.

A, B Editing outcomes of different gRNA designs for gRNA R1 and R7. C, D Sanger sequencing traces and %T values of a cell clone picked from the 15 nt wildtype sequence condition shown in (A and B). E Editing outcomes of six RUNX1 targeting gRNAs when expressed as either individual 20 nt or individual 15 nt gRNAs under the control of the CMV promoter. F, G Editing outcomes of the same six gRNAs when expressed as an array of 20 nt or 15 nt gRNAs. The order of gRNAs in array v2 in (G) is reversed from gRNA array v1 in (F). A, B, E–G show normalized mean %T values. Normalization was performed by subtracting the mean %T of the non-targeting (nt-ctrl) condition from the mean %T of the experimental condition. Source data are provided as a Source Data file.

Fig. 5. Multiplexed base editing in diverse cell lines.

A Editing outcomes mediated by BEACON1 and BEACON2 when used with a triple gRNA array targeting RUNX1. B Editing outcomes mediated by BEACON1 and BEACON2 at 16 target sites in HeLa cells transfected with a CMV promoter-driven 16x gRNA array and the respective BE system. Editing outcomes in HEK293-B2 cells mediated by BEACON2 transfected with the same 16 × gRNA array as presented in Fig. 2F. A, B show normalized mean %T values at the highest edited C for each gRNA as determined based on the single gRNA condition. Normalization was performed by subtracting the mean %T of the non-targeting (nt-ctrl) condition from the mean %T of the experimental condition. Source data are provided as a Source Data file.