There’s always been a lot of excitement (justifiably) about the prospect of treating human disease back at the DNA/RNA level. There was a burst of enthusiasm back in the early 1990s about the prospect of antisense DNA, and coming up on 2010 or so there was a similar outpouring of hope (and hype) around RNA interference. A few years after that came a rush of interest in other mRNA-based possibilities (with the founding of Moderna as a key event in this space), and of course the recent success of the mRNA vaccination platform has attracted a great deal of attention and investment.
It’s not hard to see why there’s all this interest. Our genes are read off into mRNA, which is used as the code to produce proteins, and virtually all our disease targets in this business are proteins. Our small-molecule drugs generally go in and interfere with the function of those proteins in some way, when we can tie such effects to a hypothesis for disease. What if we could just pick those individual proteins off back at the transcription or translation stages? We wouldn’t have to go through the labor of finding individual ligands for each new target – screening, hit-to-lead, lead optimization for pharmacokinetics and off-target toxicity, all the long slog. All proteins are expressed in basically the same way, when you get down to it: if you find a way to step into that process selectively, then that should work over and over for whatever protein targets you choose. Instead of laboriously picking complicated locks year after year, it’s as if you opened up a huge storage cabinet and found it lined top to bottom with ready-made labeled keys. Just choose the ones you want. That’s the dream.
For example, there’s a gene called ANGPTL3, which codes for a protein called angiopoietin-like-3. As the name implies, it’s part of the angiopoietin family, a group of important signaling proteins in the cardiovascular system. Many of them are vascular growth factors, but angiopoietin-like-3 seems to be heavily involved in lipid processing and lipoprotein formation. It’s an inhibitor of two important enzymes in this area, lipoprotein lipase and endothelial lipase, and it’s produced exclusively in the liver. If you inhibit it in mice by either doing a genetic knockout or by dosing them with a targeted antibody, the rodents show lower fatty acid and triglyceride levels, and they’re resistant to the development of atherosclerosis. That’s very nice to see, but even stronger evidence is available (the strongest we’ve got for ideas like this): there are a few scattered human beings who have been found with loss-of-function mutations in this gene, and they show the same phenotype, and with no apparent ill effects. This leads to a condition called familial combined hypolipidemia (FHBL2), and the people with these mutations have much lower levels of all lipoproteins, lower blood cholesterol, lower triglycerides, and are at lower risk for developing atherosclerosis.
So this would seem to be a solid high-priority target for some therapeutic application at the DNA or RNA level. You already have proof that humans can get along with impaired function of the gene, and in fact they appear to be healthier, and you have a very strong mechanistic connection to explain how that happens. It is no surprise that this has been followed up on. The leading antisense company (Ionis, who have been in this space for many years) developed such an agent for this gene, and signed a collaboration with Pfizer for clinical trials. Initial human results were good: the endpoint of blood triglyceride lowering was met, with no safety signals. In November 2020, a Phase IIb began in patients with elevated triglycerides and non-HDL cholesterol lipoproteins ,and last November top-line results were announced.
The compound, now named vupanorsen, met its endpoints. It lowered the levels of angiopoietin-like-3 in a dose-dependent fashion, lowered triglycerides, and lowered non-HDL lipoproteins at all doses. So far, that’s exactly what you would have expected from the mouse results and from the “human knockouts” with natural inactivating mutations in the gene. But there was more. Treatment was also associated with elevated liver enzymes (ALT and AST), which is often a sign of hepatotoxicity. This didn’t get as far as Hy’s law, which is good, since that’s used to catch patient who are at risks of fatal liver trouble, but it was concerning. And MRI imaging of the livers of patients showed an increase in hepatic fat at “certain doses”, which also makes you wonder immediately if that’s all part of the same story. These results account for the non-triumphant nature of that press release at the time – instead of trumpeting that vupanorsen nailed its clinical endpoints, the headline of the PR is just that Pfizer is announcing the results, and a bold take-home line above the main text already says that Pfizer is continuing to review the results to determine the next steps.
Anyone who’s been around this industry even a little while knows that a statement like that is the shadow of the buzzard’s wing. And indeed, Pfizer has just announced today that they are discontinuing development and returning the rights to vupanorsen to Ionis. We learn from this press release that it was (as expected) the higher doses of the compound that were associated with liver enzyme elevation, and we also learn that although those lipid endpoints were met, the effects did not seem to be large enough to justify continued development. “Particularly after signs of liver tox”, everyone reading the press release says while reading that part.
Step back and look at this experience. ANGPTL3 is a target that’s about as well-validated as we get in this business. It is definitely associated with the exact sort of therapeutic effect that we would want for a major disease, and we have human genetic data to prove it with animal models that seem to track this very closely. What else do you want? We have the company that knows more about antisense than any other in the world (you’d have to think) partnering with one of the biggest and most experienced drug companies in the world, one with a long history in cardiovascular clinical trials. And what happens? The compound works, but not well enough to be of interest as a drug. And at the same time, it sets off red warning sirens in the liver. One of the very last things you would want to do to a patient with high lipoproteins is to give them drug-induced fatty liver disease, and the end of this collaboration surely comes as no surprise to anyone who’s been following the story.
Let’s go back to that vision of opening up the master key cabinet. That’s not what’s going on; that’s never what’s really going on. You have a big collection of keys, for sure, but when you go off and try them out in the locks they’re labeled for, it turns out that some of them open other things at the same time, stuff that you probably didn’t want opened (like liver damage). Some of them break off in the lock every time you try them. Some of them give you a nasty electric shock as soon as you turn them. And some of them open up just like you figured, and you find just what’s on the label once you turn the key, but it turns out that the contents still don’t do what you thought that they did, or not as much as you’d hoped they would. Once again, there are no shortcuts to the easy wins. You have to take your best shots, run the trials, and see what happens. And they’re not all going to work.
Totally off-topic trivial aside: that “buzzard’s-wing” link above (as those brave enough to click it have already found out) takes you to an old Roger Miller song on the subject. The most famous version of that, if that’s the adjective that applies, is probably by Charlie Louvin. I mention that because the used Toyota I drove in grad school, a white 1980 Supra, was formerly his! That’s a fact I have used over the years to amaze people who know old country and gospel music singers, which admittedly has been a smaller and smaller cohort as the years go on. . .
