Following infection by exogenous DNA, some bacteria and archaea use the CRISPR/Cas system (short for Clustered Regularly Interspersed Short Palindromic Repeats (CRISPR)/CRISPR-associated (Cas)) to copy a segment of the invading sequence into a specific region of their chromosomes, thereby archiving knowledge of the infection for the cell and all its progeny. Upon subsequent invasion with a new bit of foreign nucleic acid, the cell can refer back to that archive to see if the infection is one it's seen before, at which point the cell destroys the incoming DNA. “These repetitive loci serve as molecular ‘vaccination cards’ by maintaining a genetic record of prior encounters with foreign transgressors,” Doudna wrote in a review appearing months before the seminal 2012 Science paper (2).
Led by Doudna's postdoc Martin Jinek and researchers in Charpentier's lab, that paper demonstrated that Streptococcus pyogenes protein Cas9 is an RNA-guided DNA endonuclease, and furthermore, that Cas takes its marching orders from a short guide RNA molecule called a crRNA, plus a second required (but non-targeting) RNA called tracrRNA. The team then showed how the system could be modified: First, they combined the crRNA and tracrRNA into a single chimeric molecule, thereby making the method more user-friendly; then, by supplying a custom crRNA, they targeted the system to other DNAs, in this case, to a plasmid-bound copy of GFP in vitro.
Within months, teams led by Doudna, Joung, Harvard geneticist George Church, Feng Zhang at the Broad Institute of MIT and Harvard, Jin-Soo Kim at Seoul National University in South Korea, and Luciano Marraffini at the Rockefeller University, reported adapting Charpentier and Doudna's modified system for genome editing in cultured cells or in vivo.
According to Doudna, the system is fast, easily programmed, inexpensive, and can target essentially any 20-base sequence so long as it is immediately followed by a 3’-NGG (the so-called “protospacer-adjacent motif”). “This is probably why the six papers came out as quickly as they did,” she says. “[The system] didn't require a lot of additional fiddling or optimization.”
Church applied bacterial CRISPR/Cas to human cells in culture, using the system and homologous recombination to restore activity to an inactivated genome-integrated GFP gene and to insert a selectable market at an “endogenous” viral locus on chromosome 19. In one experiment in HEK 293T cells, the team directly compared TALENs with CRISPR/Cas9, finding that the RNA-based system was about 20 times more efficient at inducing homology-directed repair (8% efficiency vs 0.37%). CRISPR/Cas-directed NHEJ was up to 38% efficient in K562 cells, but much less so in induced pluripotent stem cells (<4%).
Naturally, Church is trying to optimize the process. But already he imagines new applications in human gene therapy, for instance to change multiple copies of specific histocompatibility genes. “If [the process] really is 20 times better than zinc fingers and TALENs, then some human gene therapies that are hurting for extra efficiency could now be conceivable,” he says.
Church's team also demonstrated one feature of the CRISPR system that isn't nearly as easy with ZFNs or TALENs: multiplexing. Because the system relies on a single nuclease guided by short guide RNAs, CRISPR/Cas can theoretically generate multiple genome edits simultaneously simply by adding multiple guide RNAs. In this case, the team showed that introduction of two guide RNAs resulted in “high-efficiency” deletion of the intervening sequence, “demonstrating that multiplexed editing of genomic loci is feasible using this approach.” (3)
Not to be outdone, Joung's team took CRISPR/Cas in vivo, targeting 10 zebrafish genes by NHEJ in embryos. Though 2 of the targeting reactions failed, the remaining 8 produced mutations at rates of from 24% to 59%.Targeting the future
One unresolved question for all genome-editing systems, especially CRISPR/Cas, is what happens beyond the targeted sequence—so called off-target effects. “There are still open questions about specificity on all these different platforms,” Joung says.
In 2011, Joung coauthored one of two studies examining ZFN specificity. Though both studies used the same ZFNs and the same target (CCR5), the resulting lists of off-targets did not overlap, says Joung, “Which implies to me that neither one is capturing the full universe of off-targets, because they're missing the ones found by the other method.”