The media reaction to prime editing was extraordinary, even overshadowing Google’s claim of “quantum supremacy” published the same week. Commentators and journalists gushed about this gorgeous new “CRISPR 3.0” technology. The breakthrough even caught Elon Musk’s attention, who retweeted a New Scientist story. Urnov was much in demand, obligingly dashing off a different analogy for each reporter who called. For Scientific American, prime editing was like a new breed of dog. For STAT, it was a new superhero joining the Avengers. For Genetic Engineering & Biotechnology News, it was a college sports star preparing to join the professional leagues. “We all hope of course it will be like Alex Morgan or Aaron Rodgers in this regard—and we should know soon.”27
While Urnov speculated that prime editing could be part of an immunotherapy clinical program within a couple of years, a few commentators took issue with Liu’s eyebrow-raising estimate of 89 percent mutations that were potentially fixable. Liu is a fastidious scientist who is not inclined to hype his results—because he doesn’t need to. In a subsequent talk a month later in Barcelona, he politely but firmly pushed back. “This is a smart audience,” he told 1,500 gene therapy experts. “You know the difference between correcting a mutation and actually treating patients.”28
But how could prime editing be used therapeutically? With protein and RNA components made up of thousands of atoms, the prime editing molecular machinery is too bulky to fit into the standard AAV vector. But Anzalone was able to use a lentivirus to perform prime editing in mouse cortical neurons. The scarcity of PAM sites is not a big issue because of prime editing’s greater flexibility with regard to the edit location. The prime editing window is much longer than traditional Cas9 editing, and the system has a lower rate of off-target effects. The system appeared safer than Cas9, for reasons that make sense: whereas CRISPR-Cas9 has just one base-pairing event (when the guide RNA aligns with the target sequence), prime editing has two additional pairing steps—the binding of the pegRNA to the target site and the pairing of the flap to the original site—which offer additional opportunities to reject an off-target sequence. “If any of those three pairing events fail, prime editing can’t proceed,” Liu said.
As with the earlier CRISPR genome editors, delivery will be a big challenge, but Liu was confident he could deliver prime editors into animals, for example by using a pair of AAVs (as used in the progeria mouse model). On the safety question, Liu stressed that all genome editors have off-target effects—chemical binding is an imperfect process, just as all prescription drugs have some sort of side or off-target effect. “Each platform has complementary strengths. All will have roles in basic research and therapeutics,” Liu said, mindful that prime editing could steal some thunder from his earlier platform companies, Editas Medicine and Beam Therapeutics. Indeed, by the time prime editing was announced, Liu’s latest company, Prime Medicine, had already been formed on paper, with funding from a Google fund and F-Prime, with some rights licensed to Beam.29
Prime editing won’t be the last word in genome editing technology. I could mention Homology Medicines, which is patching in a full gene delivered by a virus to genomic targets to treat phenylketonuria without CRISPR. Or an Israeli start-up called TargetGene Biotechnologies, which modestly claims it is developing “the world’s best therapeutic genome editing platform.” Or Tessera Therapeutics which touts “gene writing” as the route “to cure thousands of diseases at their source.”30 In July 2020, Liu unveiled another impressive riff on base editing, turning a bacterial toxin into a precise gene editor that could be delivered to mitochondria to edit mtDNA. Liu’s team used TALEs rather than CRISPR as the guide.31
Whether it takes five years or fifty, it seems inevitable that we will be able to engineer bespoke variants into the genome precisely and safely.32 The prospect of rewiring the genetic code to cure deafness or diabetes, sickle-cell or schizophrenia, is getting closer all the time.33 But why stop there?
I. In this example, the C base editor targets a C:G basepair and deaminates the C to a U, resulting in a U:G mismatch. The cell’s DNA repair processes seek to repair the mismatch in one of two ways: either by switching the U back to a C; or by fixing the G to an A, thus creating a U:A, or T:A basepair. The goal was to push the system toward the latter, resulting in a C:G to T:A substitution.
II. Deamination of A actually yields inosine (I), but this is read as G.
III. Liu later found out that five groups had conducted the same initial experiment, fusing an RNA adenine deaminase, replacing our cytidine deaminase, but all five saw no editing. “No-one else made the crazy-sounding decision that we did to go ahead and evolve one.”
IV. With four bases, each in theory mutable to three other bases, there are a total of 12 possible base substitutions. The C and A base editors developed by Liu’s lab account in total for four substitutions (CBE catalyzes C-to-T and G-to-A; the ABE catalyzes the reverse substitutions, A-to-G or T-to-C), known as transitions.
V. When Anzalone googled “Cas9-RT fusion,” he learned that Cas1, which plays a role in capturing viral sequences in the CRISPR defense system, has a natural RT activity. This shows, not surprisingly, how nature uses similar concepts to prime editing but for different purposes.
CHAPTER 23 VOLITIONAL EVOLUTION
In 2007, Anne Morriss and her partner decided to start a family. The two women went to a sperm bank and chose a donor based on a few criteria (“sporty” was high on the list). Sperm banks typically only screen donors for a couple of genetic diseases—cystic fibrosis and spinal muscular atrophy—but Morriss had no reason to be alarmed. However, a few days after giving birth to a baby boy, she received a distressing phone call from a Massachusetts public health employee.
“Is your son okay? Is he still alive?”
A stricken Morriss stammered, “Yes, I think so. I just put him down for a nap.”
“Can you go and check?”
After birth, Morriss’s son Alec had received the standard Guthrie heel-stick test, in which a drop of the baby’s blood is screened for a few dozen serious genetic disorders. Against all odds, Morriss and the anonymous sperm donor were both carriers of a rare recessive genetic trait—MCAD deficiency, which affects about 1 in 17,000 people. This gene encodes an enzyme called medium-chain acyl-CoA dehydrogenase that helps the body convert fats into energy. Morriss and her partner immediately modified Alec’s diet before any serious problems ensued. A fortuitous phone call had saved their son’s life, but hospital bags remain packed by the Morriss front door in case of an emergency.
A motivated Morriss teamed up with Princeton University geneticist Lee Silver (author of Remaking Eden) to launch a next-gen diagnostics company called GenePeeks. Using Silver’s “Matchright” algorithm, GenePeeks created “digital babies” by virtually matching the DNA of the client with potential sperm donors. The client could then see which donors were predicted to have an increased risk of creating an embryo with one of hundreds of genetic disorders. These individuals could then be excluded from the donor pool, without selecting or freezing spare human embryos.