It was just a matter of time before someone mutagenized Cas9 to try to change its PAM preferences. (What's a PAM? Look here if you aren't hip yet). Although that "NGG" motif is pretty abundant in genomes - heck, it's only 2 bases - it sure would be nice to be able to target even more sequences with high specificity. For example, the closer the CRISPR site is to the site where precise genome editing is required, the more efficient it will (probably) be; having more PAM choices will only be helpful in this situation. But, having more targets isn't very useful unless the properties of CRISPR specificity and sensitivity remain robust.
So now the Joung lab has led the way with efforts to coax Cas9 into preferring new PAMs, as described in Kleinstiver et al's new paper in Nature. I really like this paper. It introduces new Cas9 variants with preferences for NGA, or NGCG PAMs. The new variants are not complicated to engineer, maintain high cleavage activity, work in vivo, and have low off-target effects similar to native Cas9.
Previously, Anders et al had published a Nature paper describing Cas9-DNA structural interaction. (Senior author of this paper was Martin Jinek, who was first author of the seminal 2012 Dounda/Charpentier Science paper). An interesting nugget in that paper was a first attempt to alter Cas9 specificity by mutating the two amino acids (Arg1333 and Arg 1335), which apparently interact directly with the two guanine bases of the PAM motif, to glutamine. This was inspired by Cas9 variants from non-S.pyogenes species which prefer A-rich PAMs and have glutamines in the homologous positions. However these changes alone couldn't make S.pyogenes' Cas9 prefer NAA instead of NGG.
Enter Kleinstiver et al. They began a systematic attempt to engineer new PAM specificities into S.pyogenes Cas9. First, they used a clever bacterial assay to measure PAM preference in which the CRISPR target is within a toxic gene. Cas9 variants with different codon changes were then introduced. In this setup, target cleavage disrupts the gene and allows the bugs to grow, allowing one to sequence the survivors to figure out which Cas9 variants worked. In this manner they identified combinations of codon changes that allowed Cas9 to recognize a NGA PAM. 2 variant combos, "VQR" and "EQR" , emerged as being best at now preferring NGA over NGG.
Then, they used a different assay to measure preference for all the possible different PAMs for selected Cas9 mutants. See Figs. 1e, 1f for these data nicely visualized. For example you can clearly see how wild type Cas9 greatly prefers NGG, but has a little ability to use NAG as had been previously reported by many - in fact many off-target analyses consider NAG PAMs as well as NGGs. Then, they tested the VQR and EQR variants, which revealed that these now are sensitive to the fourth base in the PAM. Specifically, Cas9-VQR "likes" NGAG, NGAA, NGAT, NGCG the best. Interestingly, Cas9-EQR preferred NGAG almost exclusively. The authors concluded that the T1337R variant is a gain-of-function allowing sensitivity to the fourth base, which is then specified by other codon variants. Cool.
Next, they found that the "VRER" combination allowed specific preference for a NGCG PAM. Note that this GCG motif is much less common in mammalian DNA than the other PAMs - after all, it's got a CG in it - but that also means it's off-target potential is lower. Since I know lots of genes with GC-rich regions I'm betting this PAM will come in handy.
A few more points from the paper: The new variants work in zebrafish in human cells, and have good activity and low off-target effects. Additionally, they noticed that the D1135E variant actually increased PAM specificity for the wild-type NGG PAM relative to NAG - see Fig 3a, and furthermore it reduced off-target cleavage on other off-target sites that have mismatches in protospacer but (presumably) the NGG PAM. In other words D1135E reduces off-target cleavage in general (at least somewhat) without reducing on-target cleavage. Sounds good to me!
Finally, they examined two Cas9 genes from other bacterial species and showed they could carefully measure their normal spectra of PAM preferences (which are different from the NGG of S.pyogenes). In other words they are poised to do the same mutagenesis work on these other Cas9 proteins, which will add even more PAM choices to the toolkit.
i was about to write "these Cas9 variants should be widely available soon", then I thought Hmm, better check Addgene. Sure enough: VQR, EQR, and VRER expression plasmids are already available! Kudos to Keith Joung and his lab for making these available to the world. Happy CRISPRing with new PAMs!