A positive, growth-based PAM screen identifies noncanonical motifs recognized by the S. pyogenes Cas9

Abstract

CRISPR technologies have overwhelmingly relied on the Streptococcus pyogenes Cas9 (SpyCas9), with its consensus NGG and less preferred NAG and NGA protospacer-adjacent motifs (PAMs). Here, we report that SpyCas9 also recognizes sequences within an N(A/C/T)GG motif. These sequences were identified on the basis of preferential enrichment in a growth-based screen in Escherichia coli. DNA binding, cleavage, and editing assays in bacteria and human cells validated recognition, with activities paralleling those for NAG(A/C/T) PAMs and dependent on the first two PAM positions. Molecular-dynamics simulations and plasmid-clearance assays with mismatch-intolerant variants supported induced-fit recognition of an extended PAM by SpyCas9 rather than recognition of NGG with a bulged R-loop. Last, the editing location for SpyCas9-derived base editors could be shifted by one nucleotide by selecting between (C/T)GG and adjacent N(C/T)GG PAMs. SpyCas9 and its enhanced variants thus recognize a larger repertoire of PAMs, with implications for precise editing, off-target predictions, and CRISPR-based immunity.

CRISPR technologies and the search for the PAM-free nuclease

“…For instance, the first high-throughput screen for PAMs recognized by SpyCas9 based on plasmid clearance in Escherichia coli identified NAG as a PAM, albeit with weaker recognition than NGG 22 , 23 . Subsequent work from multiple groups has shown that SpyCas9 can also weakly recognize NGA, NNGG, and a selection of other sequences 21 – 25 , reflecting a general preference for purines as well as some flexibility in the PAM gap—the distance between the target and first, defined base. While recognition can come from excess nuclease concentrations that can be readily avoided 24 , many of these sequences were identified and validated under setups reflecting practical applications of CRISPR technologies, such as plasmid clearance in bacteria, DNA binding for gene regulation, or indel formation in mammalian cells 22 – 25 .…”

“…Subsequent work from multiple groups has shown that SpyCas9 can also weakly recognize NGA, NNGG, and a selection of other sequences 21 – 25 , reflecting a general preference for purines as well as some flexibility in the PAM gap—the distance between the target and first, defined base. While recognition can come from excess nuclease concentrations that can be readily avoided 24 , many of these sequences were identified and validated under setups reflecting practical applications of CRISPR technologies, such as plasmid clearance in bacteria, DNA binding for gene regulation, or indel formation in mammalian cells 22 – 25 . High-throughput screening indicated that virtually all of the originally characterized Cas9 nucleases also recognize less-preferred PAMs 20 – 23 , 25 , representing a common theme for CRISPR nucleases.…”

“…While recognition can come from excess nuclease concentrations that can be readily avoided 24 , many of these sequences were identified and validated under setups reflecting practical applications of CRISPR technologies, such as plasmid clearance in bacteria, DNA binding for gene regulation, or indel formation in mammalian cells 22 – 25 . High-throughput screening indicated that virtually all of the originally characterized Cas9 nucleases also recognize less-preferred PAMs 20 – 23 , 25 , representing a common theme for CRISPR nucleases. These studies underscore that PAMs are not solely a consensus sequence or a motif and instead represent a landscape of sequences with different extents of recognition.…”

“…As most of these motifs were at least partially recognized by the WT SpyCas9, the end result was reshaping rather than recreating the PAM profile. The researchers also isolated a variant (D1135E) that exhibited reduced recognition of the less-preferred PAMs NGA and NAG, although a separate study showed that this variant still recognized other less-preferred PAMs like NNGG 25 .…”

“…In vivo methods based on target cleavage have relied on three approaches: clearance of a plasmid encoding the target and PAM library in bacteria 22 , selecting gRNAs that target along the genome of an infecting RNA phage in bacteria 112 or evaluating editing frequencies using constructs encoding both the guide RNA and target in human cells 113 . Separately, two different methods in bacteria link target binding by a catalytically dead nuclease to green fluorescent protein fluorescence or growth that both enrich for recognized PAM sequences 25 , 46 . Finally, cell-free transcription–translation systems (TXTL) offer a more rapid and scalable means to determine PAM sequences by eliminating cell transformation and growth as well as protein and RNA purification 50 .…”