Harnessing a diverse collection of CRISPR-associated RNA-guided nucleases and novel reverse transcriptases for precise gene editing​

My name is Drew Kelso, and I’ll be sharing how ElevateBio’s RT editing platform helps partners identify, engineer, and optimize precise gene editing systems for their target of interest. This is especially true when PAM access editing potency and guide performance limit conventional approaches.

So at ElevateBio our platform is built off five editing modalities. This originally started with identifying new nucleases. So we built out a proprietary database and we found that we could access through our enzymes compared to others a lot more of the genome. And so when you’re treating diseases this gives you a lot more access to be able to treat more diseases.

We then decided, as we built out this database even further, that we wanted to go back to the well and look for other enzymes that we could fuse to our nucleases, such as deaminases, which are used in base editing, and reverse transcriptase, which are used in RT editing. And that’s what this poster is about.

So for RT editing, this is where you take your nucleus and you fuse a reverse transcriptase to it. And then by adding a short segment to the back of the guide, you can essentially make any edit you want. And it’s highly precise because you’re controlling almost all of it. So it is a really beneficial type of edit, and it builds upon both nuclease and base editing. What we wanted to do first was just take the field standard reverse transcriptase, which is M-MLV-RT, and we look to see if it worked in our hands.

And what we found was that in the pure primary murine hepatocytes, that we could get really good levels of editing. This is about as good as it can get. So close to 45% in mouse hepatocytes.

And we then transition that to in vivo studies with the mouse using our LNP and we found that we also got really good editing there. It was close to 30%. And this is in bulk liver which is really efficient editing. And more importantly we were able to see the therapeutic effect that we wanted with this target. We then said okay, it worked with M-MLV-RT, how about we look to see if we can mine our own reverse transcriptase? Because potentially you could get them better or different or smaller. There are various things that you would be looking at for that.

So then we looked at four different classes of enzymes in our pipeline and found that we had quite a few sequences, in the thousands. And within these, if you look at, for example, the retrotransposons, we found over 38,000 sequences identified. And then what’s shown here is editing in HEK cells with plasmid. So the editing is expected to be low

But each of those bars represents a different group and that’s based on identity. So each group we could then technically expand and do more searches and more engineering. But what we did for this project is we took one of our more promising retroviral RT’s and then decided to do engineering on that.

And so in the first panel, you can see that through our first round of engineering, we were almost two-fold greater than M-MLV-RT, which is the field standard. We then did two other types of engineering. And in both of these panels, you can see that we were able to get in the middle one about a twofold increase using directed evolution and rational design and artificial intelligence for engineering. And in the right one we were able to use other rational design techniques. And we also got about a 1.5-fold increase in editing.

The other component of RT editing, as mentioned, is the guide. And this you can do a lot to. So we added chemical modifications to our RT guides and using a few different types of highly modified guides, we were able to get almost a twofold increase at every different dose compared to our parental with and without accessory factors.

So with these gains in RT editing, efficiency, and potency, we then feel very confident that we have good enzymes that we could then go in vivo and in this example, get even higher than we did previously.

Mining a database of more than 10 billion natural proteins and engineering both the RT editor and the guides, this RT platform gives partners a broader, more flexible path to deliver genomic medicines for disease therapeutics.