The multiprotein Csm complex comprises five subunits (Csm1-5) in varying stoichiometries and relies on an additional protein, Cas6, for processing the precursor crRNA40,41,42,43,44,45,46,47 (Fig. 1b). The crRNA lies at the core of the complex, with Csm1 and Csm4 binding the 5′ end, Csm5 binding the 3′ end and multiple copies of Csm2 and Csm3 wrapping around the center. The complex contains a groove along its length into which target RNAs can enter and hybridize to the variable spacer region of the crRNA. Csm1 and Csm4 specifically recognize the 5′ region of the crRNA derived from the CRISPR repeat. Each Csm3 subunit has ribonuclease (RNase) activity, leading to multiple cleavage sites within the target RNA spaced six nucleotides (nt) apart (Fig. 1c). Csm1 functions as a nonspecific single-stranded DNase (ssDNase)48,49 and a cyclic oligoadenylate (cA) synthase50,51 (Fig. 1b). The ssDNase activity is thought to defend against actively transcribed (R-looped) or ssDNA foreign genomes48,49, while the latter acts as a second messenger that activates downstream effectors in trans, such as the RNase Csm6 (refs. 50,51). Notably, all three catalytic activities are performed by independent domains of the Csm complex and can be individually ablated.
To test this, we fused GFP to catalytically inactivated Csm3 (Fig. 4a), the most abundant Csm subunit (≥3 per complex), thereby allowing multivalent display
…
By fusing GFP to the most abundant subunit (Csm3), we were able to achieve multivalent display (≥3x GFP per complex), which may offer unique advantages over single-subunit effectors such as Cas13.
Another advantage of our system is its ease of multiplexing. Multiple spacers can be cloned into the CRISPR array and processed into individual crRNAs by Cas6. This allows for pooled screening, either by encoding crRNAs against multiple targets at once or encoding multiple crRNAs against the same target. The latter may enable robust KD on the first try without the need to individually screen multiple crRNAs against a target.
Because, like other RNA-targeting CRISPR-Cas systems, Csm does not have any PAM requirement for target site selection, the only criteria we used were that the target be a unique sequence in the human transcriptome and the spacer avoid stretches of ≥5 consecutive Ts, which might cause premature Pol III transcriptional termination within the crRNA sequence.
A more large-scale analysis must be performed to determine optimal spacer design criteria, and to test how different factors (for example, melting temperature, GC content and target site availability) influence KD efficiency.