Okay, here are my prime editing takes. I'm mostly going to try to be a voice of reason here. Not because I'm hyper critical of the paper but because it's necessary. This field is full of mega hype and cool tech, but cool doesn't get us anywhere unless it works as advertised.
Overall, I think the work is great. The paper is very rigorous (lots of sites, multiple cell types, and good controls). It certainly deserves recognition because it's a nice approach that's well-executed. That being said, everyone needs to pump the brakes a bit.
Anyone who actually works in the gene editing field (and particularly people who work in vivo) knows that there are no magic bullets and that EVERYTHING is harder when you move out of a dish. Also, more components/moving parts = more potential problems. So here we go.
1. HYPE!! I know this is mostly the Broad PR machine at work and that some things come off differently in writing than when talking to the authors, so this is not a direct shot at them. But the whole "this has the potential to cure/treat 89% of genetic diseases" needs to go away.
That statement is based purely on types of mutations and ignores everything else about these diseases and mutations. Chromosomal location, local sequence context, potential off-targets, cell type, developmental stage, etc etc etc all impact therapeutic approaches.
This is particularly important here because the paper's work in primary neurons shows much more modest editing than the cancer cell lines. Plus, even in U2OS cells the editing rates are not overwhelming, suggesting it will likely be much harder to get to work in vivo.
The cell type question is also important because how different cell types + immune cells deal w/ high levels of RT in vivo could be a significant hurdle. And different cells may also have different capacities for flap repair, which is critical for this method to work.
Additionally, while much better than vanilla Cas9, indel rates are still ~10% when making non-SNP edits. So it's an improvement, but it's not indel-free. I find the method intriguing and probably applicable to a good # of mutations, but I do not see it as some sort of panacea.
Sure, in some perfect bubble of a world where mutations exist solely as short sequences of DNA, that 89% statement is true. But we don't live in that world and it's overselling to the max to say that. But it's still an additional tool that has the potential to be very useful.
2. Off-Target Effects. @GaetanBurgio hit the nail on the head w/ the missing WGS control. While I'm sure this is coming, these data are ESSENTIAL for understanding the actual usefulness of this method in therapeutic settings.
It's possible that this RT could reverse transcribe the pegRNA and that this DNA could randomly be inserted at DSBs or endogenous nicks in a sequence independent manner. If that were the case, it's a MAJOR issue. And this is basically the type of thing that killed (IMHO) CBEs.
While this may be unlikely at single nicks since the system didn't work during their first iterations w/o the 2nd nick, it's still something that needs to be explored since it could depend on things like cell cycle status, genomic locus, cell type, context, etc.
It's also possible that the RT could interfere with native RNAs, causing transcriptional disruption and cellular dysfunction. And, again, this could occur in a cell type-dependent manner, so sooooo much more needs to be explored.
3. Delivery. Again, how the immune system deals w/ this is a big question for in vivo applications. But size is also a major problem. It's huge. And while we're getting better at delivery, it's still one of the most significant hurdles in the gene editing field.
Viral delivery has issues (too big for one virus, but delivery of split systems = decreased efficiency). Nanoparticles aren't there yet. Delivery of RNPs opens up possibility of immune recognition prior to delivery to target site. There's still just a lot to solve.
P.S. - Novelty. Look, the paper is very cool. But they're not the first people to think of this sort of thing. Yes, the paper below uses yeast and retrons, but the basic principle is exactly the same. (it was one of my favorite papers from last year!)
cell.com/cell/pdf/S0092…
cell.com/cell/pdf/S0092…
I know that not everyone is up-to-speed on the field, but that doesn't excuse overlooking prior work. And it's pretty disappointing that this paper wasn't cited in the prime editing paper. It was a Cell paper + presented last year at CSHL so it was very visible. No excuses.
tldr; while prime editing looks great in 293 cells, I'm slightly underwhelmed by the U2OS + neuronal data and in vivo applications are yet to be seen. Overall, it's important to keep in mind how hard gene editing really is before declaring anything the answer to all our problems.
More nuance is always a good thing. We all want this technology to work and for therapeutic gene editing to become safe & effective. But we need to be critical + fair, give credit where it's due, vett new tech thoroughly, and leave the mega hype to the PR departments.
I should also add s/o to @liugroup @davidrliu because the work is really very nice. And very grateful for the quick dissemination of the supplementary info + getting the plasmids on @Addgene. Already ordering them + designing our first PE experiments in embryos. Will keep posted!
Well, this blew up quite a bit more than I expected, so thanks everyone for reading and sharing. 🤯 Will try to get around to all the comments, follows, etc. ASAP. For the moment, I’ve got a revision to work on. 😬
