Engineered CRISPR-OsCas12f1 and RhCas12f1 with robust activities and expanded target range for genome editing

Six previously uncharacterized Cas12f1 proteins mined from assembled bacterial genomes are active nucleases in mammalian cells, assayed by a single-strand-annealing EGFP reporter in which a Cas12f1-induced double-strand break reconstitutes a split EGFxxFP cassette. Four of them — OsCas12f1, HsCas12f1, Cb1Cas12f1 and RhCas12f1 — give robust reporter activation, but their editing of endogenous genomic loci is modest, with indel frequencies of 1% to 20%. OsCas12f1 from Oscillibacter sp. (433 aa) and RhCas12f1 from Ruminiclostridium herbifermentans (415 aa) are the most active, and their genes are less than half the size of SpCas9, LbCas12a or SaCas9.

OsCas12f1 and RhCas12f1 recognize complementary PAM classes, which makes them jointly cover targets neither reaches alone: in vitro cleavage of a DNA library carrying a 7-bp random sequence gives 5′-YTTH (Y = C or T, H = not G) for OsCas12f1 and 5′-NCCD (D = not C) for RhCas12f1. A 20-nt spacer is optimal for both.

Cleavage by OsCas12f1 requires the conserved RuvC active-site aspartates: either D228A or D406A abolishes its activity. The equivalent substitutions in RhCas12f1, D210A or D388A, likewise inactivate that nuclease.

OsCas12f1 and RhCas12f1 cleave double-stranded DNA across a wide temperature range with a 37 °C to 50 °C preference, act on both supercoiled and linear plasmids, and cut 21 to 25 bp downstream of the 5′ PAM leaving sticky ends. Both form dimers in the presence of their sgRNA by size-exclusion chromatography, as Un1Cas12f1 does.

Substituting non-basic residues in the nucleic-acid-binding regions of a hypercompact Cas12f1 with arginine raises its activity, presumably by strengthening contacts to sgRNA and target DNA. Screening over 100 single-arginine variants of OsCas12f1 across three predicted binding regions identified D52R as best at 1.31-fold over wild type, and saturation mutagenesis at position 52 confirmed arginine as the optimal substitution there. A second round combining D52R with other hits found further gains from arginine at A54, S119, T132 and S141, with T132R/D52R the most active double mutant.

sgRNA stem stability, not stem length, is what limits Cas12f1 activity. Truncating base pairs in the RNA stem region of the OsCas12f1 sgRNA impaired rather than improved cleavage, whereas replacing A-U or mismatched base pairs with thermodynamically stable C-G pairs substantially raised activity — the Os-sg1.1 variant, carrying A-U to C-G substitutions in tracrRNA stem 1, gave a 1.56-fold increase over wild type. A second round substituting further C-G pairs on Os-sg1.1 yielded 8 of 13 variants better still, with Os-sg2.6 best.

Protein engineering and sgRNA engineering of Cas12f1 are additive. Pairing the T132R/D52R protein variant with the Os-sg2.6 sgRNA produces enOsCas12f1, which exceeds either component alone and reaches 9.4-fold the activity of wild-type OsCas12f1 at the DMD locus. enRhCas12f1 was built the same way, combining high-performance RhCas12f1 protein variants with the Rh-sg1.1 sgRNA.

Engineering broadens the accessible PAM as well as raising activity. enOsCas12f1 is active across 5′-NTTN sites, achieving over 10% indels at 12 of 12 5′-TTC, 7 of 9 5′-TTA, 9 of 11 5′-TTT and 4 of 11 5′-TTG sites across 44 endogenous loci, giving a preference order 5′-TTC > 5′-TTA > 5′-TTT > 5′-TTG. enRhCas12f1 recognizes 5′-CCD, with over 10% indels at 9 of 12 5′-CCA, 4 of 11 5′-CCT and 3 of 11 5′-CCG sites across 45 loci. The comparator Un1Cas12f1_ge4.1 remains confined largely to 5′-TTTR.

enOsCas12f1 substantially outperforms the previously engineered Un1Cas12f1_ge4.1 at endogenous human loci. Targeting every PAM-accessible exonic site in PCSK9, TTR and VEGFA, enOsCas12f1 produced over 1% indels at all 30 tested sites with a maximum of 96.2%, and enRhCas12f1 at 53 of 61 sites with a maximum of 93.3%, while Un1Cas12f1_ge4.1 reached 22 of 27 sites with a maximum of 60.6%. Average efficiencies were 54.7 ± 29.8% for enOsCas12f1 and 23.3 ± 26.8% for enRhCas12f1 against 14.0 ± 18.1% for Un1Cas12f1_ge4.1 — 3.9-fold and 1.7-fold higher. With identical sgRNAs at PCSK9 and TTR, enOsCas12f1 gave 78.6-fold higher indel frequency at 5′-TTC PAM sites and 8.4-fold higher activity when each nuclease used its own preferred PAM. enOsCas12f1 also outperformed the engineered SpCas9 variant SpG at shared 5′-TTC-N20-3′-NGN sites.

enOsCas12f1 and enRhCas12f1 repair outcomes are dominated by deletions that alter the protospacer rather than insertions, and the centre of the deletion sits in the PAM-distal region outside the protospacer — the same pattern seen with Un1Cas12f1 and AsCas12f1.

The two engineered Cas12f1 nucleases discriminate mismatches in opposite parts of the spacer. enOsCas12f1 does not tolerate a single mismatch at position 3, 5 or 11 while mismatches elsewhere only slightly reduce editing, and adjacent double mismatches anywhere in positions 1 to 16 substantially reduce its activity. enRhCas12f1 partly tolerates PAM-distal mismatches, particularly at positions 19 and 20, but mismatches close to the PAM sharply reduce activity.

Genome-wide, enOsCas12f1 and enRhCas12f1 edit as specifically as established nucleases. Targeted deep sequencing at in-silico predicted off-target sites shows on-target efficiency comparable to LbCas12a and slightly above Un1Cas12f1_ge4.1 with strikingly low off-target editing, the one exception being a low off-target signal at the CLIC4 OT7 site for enOsCas12f1. By PEM-seq at the target 36 site, enOsCas12f1 gave a 7.03% translocation rate against 8.44% for Un1Cas12f1_ge4.1, 9.22% for LbCas12a and 8.19% for SpCas9, and enRhCas12f1 produced no detectable off-target site at PCSK9 where SpCas9 produced two.

Fusing the destabilizing domain of E. coli dihydrofolate reductase to enOsCas12f1 makes editing small-molecule dependent: DD-enOsCas12f1 is targeted for proteasomal degradation unless trimethoprim blocks it. Because enOsCas12f1 and two sgRNAs fit in one AAV vector, a single AAV9 injected into the tibialis anterior of a humanized DMD mouse carrying a deletion of exon 52 and human exon 51 deletes exon 51 as a ~1700 bp genomic excision, restoring the dystrophin reading frame. Exon-51-deleted transcripts reach 22.7 ± 9.2% with enOsCas12f1 and 15.0 ± 7.0% with DD-enOsCas12f1, and dystrophin protein is restored in 11.6 ± 4.0% and 7.6 ± 2.4% of myofibers respectively.

Catalytically dead enOsCas12f1 carries epigenetic effectors in a package far smaller than Cas9-based equivalents. miniCRISPRoff, built on enOsCas12f1 at 1444 aa against CRISPRoff's 2361 aa, silences GFP in stably expressing HEK293T cells in three of four versions tested and heavily methylates the Snrp promoter by bisulfite sequencing. Fusing dead enOsCas12f1 to the VPR activation domain robustly activates GFP in TRE3G-GFP reporter cells.