Perfectly matched 20-nucleotide guide RNA sequences enable robust genome editing using high-fidelity SpCas9 nucleases

High-fidelity SpCas9 variants lose most of their on-target activity in plant cells when the guide carries an extra 5′ nucleotide. Across seven rice genomic sites in protoplasts, wild-type SpCas9 gave 3% to 20% indels at every site, while with 21-nucleotide sgRNA-AN20 guides SpCas9-HF1 activity was almost negligible at five of them and eSpCas9(1.0) and eSpCas9(1.1) remained far below wild type. At the two sites whose target sequence naturally begins with adenine, so that the U3 transcript yields a perfectly matched 20-nucleotide AN19 guide, all three variants reached 40% or more of wild-type activity.

It is guide length rather than mismatch that disables high-fidelity SpCas9 variants. Adding an extra adenine to the 5′ end of guides for the two sites that already yielded perfectly matched 20-nucleotide guides significantly reduced variant cleavage, including at a site where the resulting 21-nucleotide AN20 sequence was still precisely complementary to the genomic sequence.

The constraint applies to U6-driven guides as well as U3-driven ones, so it is a property of the nucleases rather than of one promoter. At a rice site beginning with G, a wheat U6:sgRNA-GN19 construct yielding a precisely matching 20-nucleotide guide let all three high-fidelity variants induce on-target changes comparable to wild-type SpCas9, while U6:sgRNA-GN20, which adds a 21st G, sharply reduced them.

Endogenous tRNA processing supplies guides of exactly the length high-fidelity SpCas9 variants require. A tRNA–sgRNA precursor transcribed from the U3 promoter is cleaved at predefined sites by cellular RNase P and RNase Z, releasing a mature sgRNA with a precise 20-nucleotide guide fully complementary to the target.

Guides released from tRNA–sgRNA precursors restore high-fidelity SpCas9 variants to wild-type-level editing in rice protoplasts. Their on-target activities greatly exceed those with 21-nucleotide guides at the five sites where the extra nucleotide had crippled them, with the gain much larger for eSpCas9(1.1) and SpCas9-HF1 than for eSpCas9(1.0); at four of those sites eSpCas9(1.1) and SpCas9-HF1 reach or exceed wild-type SpCas9 activity. Wild-type SpCas9 activity also rises with tRNA-processed guides, but far less.

Raising sgRNA abundance does not cost specificity. Synthetic tRNA–sgRNA genes yield more U3 promoter-derived sgRNA transcript than sgRNA alone, yet off-target indel frequencies with U3:tRNA–sgRNA-N20 guides are similar to those with U3:sgRNA-AN19 guides.

High-fidelity SpCas9 variants keep their specificity advantage when charged with tRNA-derived 20-nucleotide guides. Across five off-target sites — two of one target bearing one and two mismatches, three of another bearing one, two and three mismatches — all three variants show substantially less off-target editing than wild-type SpCas9, most clearly at off-targets with two or more mismatches, with SpCas9-HF1 consistently lowest. On-target to off-target indel frequency ratios average 273-fold higher for the variants than for wild-type SpCas9.

Systematically walking pairs of mismatches along a guide sequence confirms the fidelity of high-fidelity SpCas9 variants delivered through tRNA–sgRNA fusions: with each mismatched guide fused to tRNA and transcribed from the U3 promoter, eSpCas9(1.0), eSpCas9(1.1) and SpCas9-HF1 all induce much lower indel levels than wild-type SpCas9.