Engineered Cas12i2 is a versatile high-efficiency platform for therapeutic genome editing

CRISPR-Cas systems provide adaptive immunity in prokaryotes and have been adapted for genome editing. Class 2 systems (e.g., Cas9 and Cas12a) are widely used therapeutically. Recently identified type V systems include subtypes V-G, V-H, and V-I. The subtype V-I system (Cas12i) is evolutionarily distinct (clustering with V-B/Cas12b) yet functionally resembles Cas12a in that it can process pre-crRNA and does not require a tracrRNA for DNA cleavage. Cas12i effectors are smaller (approximately 1033–1093 aa) and use short mature crRNAs (40–43 nt), supporting multiplexing and viral delivery. Despite in vitro cleavage activity, Cas12i1/2 had not demonstrated therapeutic utility in mammalian cells due to low editing efficiency. This study addresses whether engineering Cas12i2 can overcome low activity to create a compact, tracr-less, efficient, and specific nuclease suitable for ex vivo and in vivo therapeutic genome editing. The authors applied high-throughput mutational scanning to enhance Cas12i2 activity and assessed editing efficiency and specificity across human cell lines, primary T cells, and CD34+ HSPCs, as well as AAV delivery performance.

- Assessed wild-type Cas12i2 (Cas12i2 WT) activity in HEK293T cells by plasmid expression of NLS-tagged effectors and U6-driven gRNAs targeting multiple loci; measured indels by targeted deep sequencing. - Semi-rational engineering: In absence of Cas12i structural data, substituted 480 residues (aa 575–1054) in the C-terminal nuclease lobe (RuvC-containing) with Arg or Gly to potentially enhance nucleic acid binding (Arg) or conformational flexibility (Gly). Generated 960 single-substitution variants via overlapping-PCR and screened using an in vitro fluorescent reporter assay coupling reconstituted E. coli cell-free protein synthesis with target-dependent depletion of GFP expression. Calculated depletion Z-scores normalized to wild-type controls; selected candidates for cellular testing. - Cell-based validation: Transfected HEK293T cells with top in vitro hits; quantified indels at two genomic sites; combined top mutations (D581R, I926R, V1030G) into ABR-001 and compared activity versus WT and SpCas9 across 18 loci. Analyzed indel size distributions. - In vitro biochemistry: Compared dsDNA cleavage efficiency of ABR-001 vs WT Cas12i2; mapped cut sites on target and non-target strands using labeled PCR amplicons, mung bean nuclease treatment, and denaturing PAGE. - Specificity profiling: Used tagmentation-based tag integration site sequencing (TTISS) to empirically identify double-strand break sites (on- and off-targets) across 18 targets in three loci, comparing ABR-001 vs SpCas9. Complemented with in silico off-target prediction (searching sequences adjacent to NTTN PAMs with up to edit distance 6), followed by targeted deep sequencing of top 10 predicted off-targets per guide and MLE correction for background. - Ex vivo editing in primary human cells: Prepared ABR-001 RNPs (ABR-001 protein complexed 2.5:1 molar with crRNA). Electroporated into stimulated human CD3+ T cells targeting B2M (dose-response 2–16 µM), TRAC, and CIITA; measured viability, protein knockdown by flow cytometry, and indels at day 7. In CD34+ HSPCs, targeted the BCL11A erythroid enhancer (three guides and a dual-guide multiplex) to disrupt the GATAA motif; assessed indels and motif disruption at 72 h, colony-forming capacity (CFC), persistence of edits and indel profiles during 20-day erythroid differentiation, and HbF induction by intracellular staining. - In vivo persistence: Transplanted ex vivo edited CD34+ HSPCs (200k cells/mouse) into irradiated NSG mice; measured indel persistence at 8 and 16 weeks, HbF expression in human CD45+ marrow cells, and engraftment/lineage markers by flow cytometry. - AAV delivery: Constructed AAV2 vectors expressing ABR-001 (CMV promoter) and a U6-driven crRNA in a single vector; transduced HEK293T at varying MOIs; quantified indels at 72 h and over 7 days; compared with SaCas9 AAV at matched targets. Tracked indel size distributions over time. - Sequencing and analysis: Prepared amplicon libraries (two-step PCR), sequenced on Illumina platforms; computed indel rates and size distributions using a k-mer scanning pipeline; for TTISS, processed reads to identify integration windows meeting PAM and cut-site proximity filters and removed events present in negatives.

- Wild-type Cas12i2 showed low mammalian editing: mean indel 1.1% across 18 HEK293T targets, indicating insufficient activity for many applications. - Engineering outcomes: Of 14 in vitro hits tested in cells, three single substitutions (D581R, I926R, V1030G) improved indels by ~1.5–2× relative to WT at one/both sites; combining them (ABR-001) yielded ≥3× improvement over WT and editing efficiencies approaching SpCas9 across 18 loci. ABR-001 edits were biased toward larger deletions (5–20 nt) compared with the small deletions/+1 insertions typical for SpCas9. - Biochemical characterization revealed increased dsDNA cleavage for ABR-001 vs WT and multiple cut sites including PAM-distal positions on both strands; extended NTS cuts were also seen with WT, indicating they are intrinsic to Cas12i2. - Specificity: TTISS detected fewer potential off-targets with ABR-001 than SpCas9 across 18 targets. Example: at VEGFA target 1, ABR-001 had no detectable off-targets, whereas SpCas9 had one major and six minor. In silico-guided deep sequencing found off-target indels (>0.2%) for 2/18 ABR-001 targets vs 8/18 for SpCas9, acknowledging the biased nature of this approach. - Ex vivo T cell editing: ABR-001 RNPs targeting B2M produced robust, dose-responsive editing with indels approaching ~90% at 16 µM and maintained >70% viability across doses; efficient protein knockdown was observed. TRAC and CIITA were also efficiently edited with maintained viability and strong protein knockdown. Indel profiles in primary T cells retained the large deletion bias. - CD34+ HSPCs: Targeting the BCL11A enhancer yielded robust indel rates comparable to SpCas9. Two guides significantly disrupted the GATAA motif, though to a lesser extent than SpCas9; dual-guide multiplexing maximized both motif disruption and overall indels. Edits persisted through 20 days of erythroid differentiation with characteristic broad deletion profiles (broader with multiplex). HbF levels in ABR-001–edited cells were high and equivalent to SpCas9 despite lower indel/motif disruption. In NSG mice, ABR-001–edited cells maintained indels up to 16 weeks and showed SpCas9-equivalent HbF in human CD45+ marrow cells; engraftment was similar or higher than controls. - AAV delivery: Single-vector AAV2 delivery of ABR-001 achieved up to ~60% indels at 72 h at the highest MOI, matching or exceeding SaCas9 at matched loci; indels increased over time to >80% by day 4 and then stabilized. Indel patterns remained target dependent and consistent after day 4. These results highlight ABR-001’s suitability for in vivo delivery via AAV.

The study addresses the challenge of low mammalian editing activity in Cas12i by semi-rational engineering to enhance DNA binding and catalytic performance. Combining three substitutions (D581R, I926R, V1030G) produced ABR-001, which delivers near-SpCas9 on-target efficiencies while maintaining high specificity characteristic of type V nucleases. Biophysical data suggest improved dsDNA binding affinity contributes to enhanced activity; structural modeling implies that D581R (WED) and I926R (Nuc) may strengthen electrostatic interactions with the dsDNA backbone near the PAM and non-template strand, respectively, while V1030G (RuvC) may positively influence catalysis. Functionally, ABR-001’s broad deletion profiles can be advantageous for disrupting non-coding regulatory elements, as evidenced by effective HbF induction via BCL11A enhancer editing despite lower motif disruption compared with SpCas9. The compact protein and tracr-less guide enable practical AAV single-vector packaging and multiplexing. Across HEK293T cells, primary T cells, and CD34+ HSPCs, ABR-001 showed robust editing, low off-target activity by TTISS and targeted assays, durable edits through differentiation and in vivo engraftment, and strong performance with AAV delivery. Together, these findings establish ABR-001 as a versatile platform for both ex vivo and in vivo therapeutic genome editing.

This work demonstrates that semi-rational engineering of Cas12i2 yields ABR-001, a compact, tracr-less type V-I nuclease with robust on-target activity, high specificity, and broad applicability for therapeutic genome editing. ABR-001 approaches SpCas9 efficiency, exhibits advantageous large-deletion profiles for non-coding element disruption, supports efficient ex vivo editing of T cells and CD34+ HSPCs with durable function and engraftment, and achieves high editing when delivered via a single AAV vector. Future efforts leveraging emerging Cas12i structural insights could further optimize activity and specificity, expand PAM compatibility, and tailor indel outcomes. Additional in vivo studies across tissues and disease models, and comprehensive off-target profiling, will advance clinical translation.

- The in vitro fluorescent reporter screen did not always predict cellular editing performance; many Arg/Gly substitutions failed to improve or even reduced activity in cells, highlighting context dependence. - Off-target evaluation via in silico prediction covered only the top predicted sites per guide; while TTISS provided unbiased detection, both methods have detection limits and filtering steps that could miss rare events. - Editing efficiencies, indel profiles, and functional outcomes were target dependent; generalization to broader genomic contexts requires further validation. - The mechanistic contribution of V1030G remains unclear due to limited local structural information. - AAV delivery was evaluated in vitro (HEK293T) rather than in animal models; in vivo efficacy, biodistribution, and safety require further study. - In vivo HbF induction in engrafted marrow was SpCas9-equivalent but not significantly elevated over untransfected controls in this study, potentially reflecting model sensitivity, sample size, or assay timing.