The Neisseria meningitidis CRISPR-Cas9 system enables specific genome editing in mammalian cells

Optimal guide length for NmeCas9 depends on whether the guide is delivered as two RNAs or one. Testing guides from 24 down to 20 nucleotides, the native two-RNA system with crRNA and tracrRNA expressed from separate U6 promoters prefers longer guides, while a chimeric single guide RNA expressed from one U6 promoter tolerates more length variation. Across configurations, NHEJ-induced indel rates range from 3% to 75% depending on locus and guide length.

Base identity at two positions modulates NmeCas9 activity independently of guide length: adenine is disfavoured relative to thymine as the first N of the ANNNGATT PAM, and activity is lower when thymine occupies the PAM-proximal base of the target sequence.

In mammalian cells NmeCas9 is most active at its bioinformatically predicted NNNNGATT PAM, with NNNNGCTT and NNNNGTTT also supporting high activity and several further PAMs giving lower activity. Measured with a luciferase single-strand-annealing reporter in HEK293T cells — where two direct repeats of the firefly luciferase open reading frame flank the target site, so cleavage and single-strand annealing restore luciferase and site accessibility differences between endogenous loci are removed — the set of permissive PAMs differs between guide formats, the two-RNA system tolerating more base variation than the single guide RNA system. These weakly active PAMs are too inefficient for deliberate targeting but must be included when searching for off-target sites.

NmeCas9 discriminates single-base guide-target mismatches better than SpCas9. Systematically substituting every position of an HPRT1-targeting guide except the 5′ guanine required as the U6 transcription start site, and comparing at each system's optimal guide length (23 nucleotides for the two-RNA Nme system, 21 for the Nme single guide RNA system, 20 for Spy), off-target cleavage from single mismatches varies by position in all three systems but SpCas9 shows higher off-target activity at most positions.

Both NmeCas9 and SpCas9 tolerate single-base DNA bulges at the guide-DNA interface, but with different positional profiles and lower overall activity for NmeCas9. SpCas9 tolerates DNA bulges at the 5′ and 3′ ends and at positions 10 to 14 from the PAM, the two-RNA Nme system tolerates none in the central region, and the Nme single guide RNA system retains significant activity with a bulge at position 6, 7 or 8. Some apparent bulge tolerance is reinterpretable as mismatches — the Nme position 6/7/8 case can be modelled as three mismatches plus a non-recognized NNNNTTTA PAM, and SpCas9 3′-end deletions shift the PAM from GGG to AGG — but tolerance of centrally located bulges at positions 10 to 14 cannot be explained that way.

Single-base RNA bulges abolish cleavage almost completely for both SpCas9 and the NmeCas9 single guide RNA system when placed outside the 5′ end, whereas the two-RNA Nme system with separate crRNA and tracrRNA tolerates RNA bulges in the central region at positions 10 to 15 — so splitting the guide changes which classes of imperfect pairing the nuclease will act on.

Comparing the two orthologs at endogenous loci requires sites carrying both PAMs, found by searching for NGGNGHTT so that the Nme NNNNGHTT and Spy NGG requirements overlap on a shared protospacer. Across eight such sites, NmeCas9 reached up to 40% indels at HPRT1, F8 and DHFR but was generally lower than SpCas9, and at the remaining five sites its activity was substantially lower.

The longer NmeCas9 PAM dramatically reduces the number of candidate off-target sites in the genome before any cleavage is measured: COSMID searches for sites with three or fewer mismatches plus one-base insertions and deletions returned far fewer candidates for Nme guides (three mismatches within the 19 PAM-proximal bases, NNNNGNTT PAM) than for Spy guides (NRG PAM).

Deep sequencing of top-ranked candidate off-target sites for guides against HPRT1, DHFR and F8 found one off-target with measurable cleavage for SpCas9 at each of the three target genes, while no measurable off-target activity was detected at any site tested for either NmeCas9 configuration. Mutation profiles differ between target loci, most likely reflecting differences in microhomology around each cleavage site.