High-fidelity Cas9 variants buy specificity by losing on-target activity or constraining guide design; OpenCRISPR-1, designed by generative AI, matched Cas9 across 28 loci while cutting off-targets up to 553-fold and tolerating every U6 guide format.
OpenCRISPR-1, a Cas9-like nuclease produced by generative AI protein design, matches canonical Cas9 on-target editing in human cells. Across 28 endogenous target sites in HEK293T cells delivered by plasmid, OpenCRISPR-1 gives slightly higher average mutation frequencies (69.73% ± 2.89%) than Cas9 (68.34% ± 3.49%), with highly similar cumulative distributions of editing efficiency across sites. Immunoblot shows comparable protein levels for the two nucleases, so activity differences reflect intrinsic enzyme properties rather than expression.
02
Relative activity of OpenCRISPR-1 and Cas9 depends on cell type, so nuclease choice cannot be generalized from one line. In human induced pluripotent stem cells OpenCRISPR-1 edits VEGFA1 more efficiently than Cas9 and matches it at EMX1, while in non-transformed MRC-5 human lung fibroblasts OpenCRISPR-1 is somewhat less active than Cas9 across the loci tested.
03
Multiplex Digenome-seq across ten guide RNAs shows OpenCRISPR-1 cleaves far fewer genome-wide candidate off-target sites than Cas9: 25.6 ± 13.3 sites against 88.5 ± 33.2 for Cas9, using recombinant protein complexed with each guide on genomic DNA followed by whole-genome sequencing.
04
OpenCRISPR-1's genome-wide cleavage is more tightly confined to canonical PAMs than Cas9's. In Digenome-seq, Cas9 cleaved sites with NGG PAMs (81.36% ± 28.37%) but also a considerable fraction with noncanonical NGH (6.10% ± 2.64%), NHG (11.75% ± 4.66%) and NHH PAMs (0.79% ± 0.77%), whereas OpenCRISPR-1 cleavage was 87.89% ± 11.50% at NGG PAMs with limited activity at noncanonical ones.
05
Targeted deep sequencing of validated off-target sites confirms large specificity gains for OpenCRISPR-1 without an on-target penalty. At the HEK2 off-target site OT1, which carries two mismatches relative to the on-target sequence, Cas9 induces 6.11% ± 0.18% indels while OpenCRISPR-1 induces 0.01% ± 0.02% — a 553-fold improvement — and OpenCRISPR-1 simultaneously maintains higher on-target activity at HEK2 than Cas9.
06
OpenCRISPR-1 suppresses off-target editing at noncanonical PAMs as well as at canonical ones. At the VEGFA2 off-target site OT2, a single-mismatch site adjacent to a non-NGG PAM, OpenCRISPR-1 indels are undetectable against 2.38% ± 0.16% for Cas9, and at the double-mismatch EMX1 off-target site OT3 OpenCRISPR-1 gives 1.90% ± 0.19% against 51.58% ± 0.71% for Cas9.
07
Delivering these nucleases as ribonucleoproteins in engineered virus-like particles raises on-target editing above plasmid transfection at VEGFA1 and EMX1 and improves specificity for both enzymes, lowering the off-target index of each. OpenCRISPR-1 delivered this way reaches up to 257-fold greater specificity than Cas9 delivered the same way.
08
OpenCRISPR-1's specificity advantage traces to PAM discrimination rather than to spacer-level differences, shown by comparing genomic loci that share an identical spacer sequence next to canonical or noncanonical PAMs. At NGG PAM sites Cas9 edits modestly better (66.62% ± 4.15%) than OpenCRISPR-1 (57.68% ± 3.63%), but OpenCRISPR-1 editing falls 86% at NHG PAMs and 70% at NGH PAMs: at NHG sites Cas9 averages 14.29% ± 6.06% against OpenCRISPR-1's 2.09% ± 1.08%, a 5.1-fold drop, and at NGH sites 6.92% ± 6.32% against 3.31% ± 3.46%, a 2.1-fold drop.
09
OpenCRISPR-1 tolerates all three guide RNA formats imposed by the U6 promoter's 5′ guanine requirement, unlike most high-fidelity Cas9 variants. With GX19 guides, whose spacer naturally begins with G, OpenCRISPR-1 matches eCas9, Cas9-HF1, HypaCas9 and Sniper-Cas9 and consistently outperforms EvoCas9 across nine endogenous loci. With gX19 guides, where a non-G first nucleotide is replaced by G, and gX20 guides, where a nontemplated G is appended to a 20-nucleotide spacer, only OpenCRISPR-1 and Sniper-Cas9 retain significantly high editing efficiency while eCas9, Cas9-HF1, HypaCas9 and EvoCas9 all lose activity.
10
OpenCRISPR-1's specificity sits within the range of purpose-engineered high-fidelity Cas9 variants and above some of them, though the ranking is locus-dependent. Cas9 shows the highest off-target index at every locus tested except FANCF. Normalized to Cas9 across eight target sites, OpenCRISPR-1's relative off-target index of 0.18 ± 0.07 is comparable to eCas9 (0.09 ± 0.04), Cas9-HF1 (0.10 ± 0.03) and HypaCas9 (0.12 ± 0.06), and better than Sniper-Cas9 (0.35 ± 0.09) and EvoCas9 (0.85 ± 0.77) — lower values indicating greater specificity. At individual sites OpenCRISPR-1's off-target index is slightly higher than eCas9, Cas9-HF1 and HypaCas9 at HEK4 but lower than all three at VEGFA2.
11
OpenCRISPR-1 converts into a functional prime editor by the same route used for Cas9: nicking-only mutation plus fusion to an engineered reverse transcriptase, giving OpenCRISPR-PE2 and OpenCRISPR-PE7 analogous to PE2max and PE7. Across three endogenous loci, OpenCRISPR-PE2 matches PE2max with canonical pegRNAs, though efficiency varies by locus and pegRNA type. Paired with the engineered epegRNAs tevo-epegRNA and mpk-epegRNA, OpenCRISPR-PE2 consistently exceeds PE2max, and OpenCRISPR-PE7 raises efficiency further above OpenCRISPR-PE2 in most conditions.
12
OpenCRISPR-based prime editors inherit the nuclease's fidelity advantage. Across seven off-target sites, five of them previously validated, both OpenCRISPR-PE2 and OpenCRISPR-PE7 induce substantially fewer off-target modifications than PE2max and PE7 respectively while maintaining high on-target editing, with the difference most pronounced when paired with optimized epegRNAs.
1 / of12
Abstract
RNA-guided nucleases such as CRISPR-Cas9 systems have revolutionized genome engineering by enabling programmable DNA modifications. Although structure-guided and evolution-derived high-fidelity Cas9 variants improve target specificity, they often compromise on-target activity or constrain guide RNA (gRNA) design. We performed head-to-head comparisons of OpenCRISPR-1 and Cas9 in human cells using amplicon sequencing, multiplex Digenome-seq, and off-target validation by targeted sequencing. Editing activity was assessed across 28 endogenous loci in HEK293T cells and further evaluated in human induced pluripotent stem cells (iPSCs) and MRC-5 fibroblasts. To test clinically relevant delivery, Cas9 and OpenCRISPR-1 ribonucleoproteins were delivered using engineered virus-like particles (eVLPs). We also generated OpenCRISPR-based prime editors, OpenCRISPR-PE2 and OpenCRISPR-PE7, and compared them with PE2max and PE7 using pegRNAs and engineered epegRNAs. Here, we show that OpenCRISPR-1, an AI-designed, Cas9-like nuclease, retains Cas9-level editing efficiency across multiple genomic loci while significantly reducing off-target mutations. Using multiplex Digenome-seq and targeted deep sequencing, OpenCRISPR-1 exhibits up to a 553-fold reduction in off-target mutations compared to Cas9 and achieves off-target indices that match or surpass those of high-fidelity Cas9 variants. OpenCRISPR-1 also sustains robust editing across diverse gRNA formats (GX19, gX19, and gX20), highlighting its enhanced versatility. Furthermore, converting OpenCRISPR-1 into a prime editor yields comparable editing efficiencies while lowering the relative specificity ratio by up to 97%. These findings establish generative AI-guided protein design as a powerful strategy to overcome the specificity-efficiency trade-off, expanding the genome editing toolkit for both research and therapeutic use, and ushering in a new era of rational protein design.
Choosing a selection results in a full page refresh.
