Zinc-finger nucleases, TALENs, and the CRISPR/Cas system are all the rage these days as tools for genome editing. But they are not the only options. There also are site-specific recombinases. Now a pair of researchers at Harvard University report mutants of one such protein, called Cre, with substantially higher site-specificity than the wild-type enzyme, a feature that makes them more attractive for human gene therapy applications.
Nikolai Eroshenko, a graduate student in the Harvard School of Engineering and Applied Sciences, and his graduate advisor George Church of Harvard Medical School, reported the findings Sept. 23 in Nature Communications. 
Eroshenko’s interest is human gene therapy, one common form of which involves replacing a mutant gene with a wild-type or “correct” copy. But inserting that copy into the genome isn’t easy. Researchers can use viruses, but cannot control the site of integration. Nucleases that introduce double-stranded DNA breaks are another option, but they require repair by homologous recombination, a relatively infrequent event in human cells.
Recombinases like Cre are highly efficient in human cells and require no endogenous DNA repair pathways. To use them, researchers must engineer recombinase binding sites (called loxP in the case of Cre) in the genome and the DNA to be inserted, and the enzyme takes care of the rest. “It’s a single-enzyme system that is able to do this reaction by itself without endogenous machinery,” Eroshenko said.
Typically, the recombinase process is thought to be quite specific. But some evidence suggested the potential for off-target effects, a problem for any potential applications in humans.
“Was there a way to push the specificity, which was already high, higher?” he asked.
To find out, Eroshenko used mathematical modeling to identify a strategy for improving the protein’s accuracy without sacrificing efficiency, and determined that mutations targeting Cre’s ability to form homodimers (as opposed to binding DNA) might do the trick.
“What the model shows is if you make the two [monomer] binding events more independent of each other, you get an increase in binding accuracy,” he explained.
Eroshenko then used PCR-based mutagenesis to target an alpha helix near the protein N-terminus that is involved in dimerization but not DNA binding. Using a double selection strategy, he identified three mutants, all of which retained the ability to specifically and correctly target the loxP site while avoiding loxP-like pseudo-sequences. When he compared those mutants to the wild-type Cre recombinase in vitro and in bacterial and mammalian cells, he found that the enzyme was slightly less efficient, but far more specific.
In E. coli, Eroshenko said, he observed an error rate with the wild type enzyme of about 1 in 10,000 cells. “We got three or more orders of magnitude better than with some mutants.”
Eroshenko says his long-term goal is to “re-engineer” the recombinases to target sites other than loxP, such as endogenous pseudo-loxP sites already in the human genome, so recombinases can be used on unmodified genomes.
Karl Clark, Assistant Professor of Biochemistry and Molecular Biology at the Mayo Clinic in Rochester, Minnesota, who studies genome engineering tools including recombinases like Cre, said he was surprised by the results. “I don’t hear a lot of people complaining about [Cre/]loxP being nonselective,” he said.
Clark said he plans to try the new mutants in his own lab to see if they work well for zebrafish in vivo. He notes that for many applications – for instance, transgenic applications in which Cre is pulsed for a short time to activate a silenced gene -- the accuracy of the wild-type enzyme is already sufficient.
That said, “there’s always a teeter-totter balance between efficiency and accuracy” with enzymes like this, and anything that can shift the balance towards greater accuracy with limited loss of efficiency is a good thing.
“For human gene therapy, most want to err on the side of accuracy.”
 N. Eroshenko, G.M. Church, “Mutants of Cre recombinase with improved accuracy,” Nat Commun, 4:2509, 2013.