NEW YORK (GenomeWeb) – Scientists from North Carolina State University have developed a method to screen for protospacer-adjacent motif (PAM) preferences in CRISPR systems that have not yet been characterized, and their results suggest that the current mode of thinking about PAMs is far too simplistic.
Led by first author Ryan Leenay and co-senior authors Chase Beisel and Rodolphe Barrangou, the scientists developed a method called PAM-SCANR, which exposes pre-complexed CRISPR ribonucleoproteins to a combinatorial library of Escherichia coli containing a genetic circuit with a target tied to a fluorescent reporter. Each cell in the library contained a unique potential PAM sequence next to the target sequence. If the ribonucleoprotein targeting the genetic circuit recognized that specific PAM, it would induce the cell to fluoresce. The scientists then used fluorescence-activated cell sorting to find the cells with PAMs recognized by the CRISPR system and those bacteria were sequenced to reveal the PAMs that were enriched for.
When the researchers applied PAM-SCANR to well-studied CRISPR systems, such as the CRISPR/Cas9 system of S. pyogenes, they found a much more complex story than has been told up to now.
The new study, published today in Molecular Cell, "brings a certain richness to what you'd describe as a PAM," Beisel said.
"When we first discovered PAMs nearly a decade ago, we initially thought that only one PAM worked," Barrangou said in a statement. "However, our tools revealed there can be multiple PAMs for a single CRISPR/Cas system, and some PAMs clearly performed better than others as part of CRISPR recognizing its target DNA."
The PAM site is essential for CRISPR systems, an initial toehold for the guide RNA to lock on to and hybridize to the genomic target, bringing Cas9 or other CRISPR-associated proteins into position to do whatever they've been designed to do, whether it's cleaving DNA or some other engineered function.
But PAM sequences differ for each of the myriad CRISPR systems evolved in bacteria and archaea. While the PAM for Streptococcus pyogenes — the current gold standard for CRISPR/Cas9 technology — has been canonically established as "NGG," other CRISPR systems are not well understood.
"There are thousands of potential CRISPR tools out there," Beisel said. "To make use of them we need an efficient way to identify their PAMs — and we think we've developed tools to do that."
To demonstrate their method, the scientists tested PAM-SCANR in several important CRISPR systems covering the diversity across bacteria and archaea, including E. coli, Streptococcus thermophilus, Francisella novicida — which exhibits the CRISPR/Cpf1 system, and Bacillus halodurans — whose PAM preferences had not yet been characterized.
In every case, the authors found multiple possible PAM sites.
"PAMs are often viewed as a single entity," Beisel told GenomeWeb. While a canonical PAM, such as the "NGG" of S. pyogenes Cas9 (spCas9) can account for variation, where "N" is any nucleotide, it doesn't capture the full complexity. While researchers had already known spCas9 can also weakly recognize an "NAG" PAM, this interaction wasn't fully understood and is basically ignored. "What a PAM is is much more nuanced and doesn't blend to this consensus that people try to come up with," he said.
"Rather than being black or white, correct or not, there's a gray zone. Some can still function even if they're not better than others," he said. "Some things work really well that don't look like the others at all."
The authors also introduced a graphical representation of PAM-SCANR data called a PAM wheel, which they hope will help make clear their points about complexity in PAM preferences.
The PAM wheel is a visual representation of the individual sequences in the library and their enrichment scores from the screen, Beisel said, and resembles a set of nested donuts. It captures all the possible PAM sequences recognized by the cell sorting and sequencing as well as the relative enrichment in the screen for each PAM.
The authors applied PAM wheel analysis to existing data sets of S. pyogenes and Staphylococcus aureus CRISPR activity. For S. pyogenes, the PAM wheel suggests there are actually two potential patterns for PAM sites: "NGGN" and "NNGG."
"You see PAMs where the fourth position is often a "G", and those are what you'd consider to be less good PAMs," Beisel said. "That suggests there is flexibility in the spacing between the target sequence and the PAM."
While the PAM analysis was only performed in bacteria and not in eukaryotic cells, where many exciting CRISPR applications take place, Beisel said he expected the results to translate and could have major implications for target selection and how to search for and mitigate off-target effects.
"This is a big caveat: we didn't do genome editing in mammalian cells. But if the results translate, how we go about determining what an off-target site looks like becomes much more complicated," he said.