Of the discovery of GSK3 (glycogen synthase kinase 3) inhibitors there has been no end. I first came across it as a target it about 1997, and even then, once I started reading the literature, I quickly felt as if I were late to the party. It’s been investigated for diabetes (and other metabolic diseases), Alzheimer’s, bipolar disorder, inflammation, various forms of cancer, and many other things besides.
A bit of background for those outside the field: kinase enzymes are a large and important class of enzymes, because their job is to phosphorylate (attach phosphate groups to) various surfaces of other proteins. Phosphorylation has evolved as a key signaling pathway; a phosphate is small, very strongly charged, and thus can be easily recognized. The number and location of these phosphates often act as switches or modulators for protein function, and can turn a given protein’s activity completely on or off, change their binding partners or substrates, change where they end up going in the cell, etc. So if you target kinase enzymes, you target an awful lot of stuff downstream, which is good, but can also be a problem. Up into the 1990s kinases were thought to be more or less undruggable, because their active sites were thought to be too similar (they all bind the ubiquitous ATP as a source of phosphate). But this has proven not to be true, and how, and the number of small-molecule kinase inhibitors described in the literature is by now beyond counting. The big questions, though, are always how selective these drug candidates are, and how selective you need them to be for a given task.
How does one enzyme get to be so popular? By doing an awful lot of stuff in the cell. GSK3 turns out to have over one hundred different substrates, and anyone who thinks that list is complete is surely mistaken. Kinases in general tend to have a lot on their agenda – here’s a study that looked into predicting how many different substrates a given kinase enzyme could theoretically have, based on what we know about the structures of the group they prefer. The authors looked at 68 different kinases in detail, and the average number of substrates predicted for each was 70. The beta-isoform of GSK3, though, had over five hundred predicted substrates, so yeah, it’s a real Swiss army knife.
The link preceding that one goes into many interesting details – the enzyme seems to be constituitively active (no off switch), and is affected by (as you’d expect) a wide variety of regulatory pathways. It also has to have an unusual active site that’s able to accommodate a number of different targets, and from an evolutionary standpoint, it’s probably just found itself being used for more and more things as time has gone on. It will be appreciated that all this fascinating complexity has made drug development against GSK3 rather fraught. That Wikipedia link in the first paragraph lists a few dozen known inhibitors, and that isn’t anywhere near comprehensive. You can buy whole lists of compounds from the biochemical suppliers, and if you look under the hood into the patent literature on the subject, you’d better stand back.
But with all this work, over all these years, I’m not aware of anyone who’s made it through the clinic with a GSK3-targeted therapy. No doubt some of the existing kinase inhibitors have activity against it, but specific efforts against the enzyme have never made it through, presumably because of the tendency for GSK3-active compounds to also hit other things, and for the inhibition of GSK3 itself to do more than you were planning on. That brings up this new paper in Science Signaling, from a group of Israeli researchers. They’ve been working on short peptidic inhibitors of the enzyme, basing it off what appear to be some of its more specific preferences, and have come up with an ingenious mechanism.
They have an 11-mer (designated L807) that is also myristoylated (functionalized with myristic acid, a 14-carbon fatty acid). That’s yet another way that proteins can be modified to change their interactions with other proteins, their intracellular targeting and location, etc., and in this case it makes the original peptide much more potent against the enzyme. But it’s not a straightforward substrate competition per se – close study of the kinetics and the other details of the reaction indicate that L807 binds to GSK3 as a substrate, and is then phosphorylated by it. This phosphorylated peptide, though, turns out to be a much tighter binder, and is the real inhibitor: the original peptide is just an odd sort of prodrug.
This “substrate-inhibitor conversion” mechanism has some advantages, selectivity being a big one. The real inhibitor is only made at each individual GSK3 active site. As long as the unphosphorylated peptide isn’t a red-hot kinase inhibitor (which it doesn’t seem to be), and as long as it’s not a great substrate for other enzymes (and especially as long as the resulting phosphorylated species, if formed, doesn’t turn out to be a great inhibitor for those other enzymes as well), then you have a potentially useful path. And this seems to be the case – the L807myr compound looks pretty clean across a diverse panel of other kinases. The whole process reminds me of Jonathan Ellman’s “substrate activity screening” idea, only this one is happening without human intervention.
Of course, if you’re proposing an 11-mer peptide derivative as a useful compound, you have to ask if it does anything when you move beyond enzyme preps and into cells and animals. Interestingly, this one does. It’s active in cell culture, and even in mice. They show results for doses of 20 and 80 micrograms in mice (converting at the standard 25g/mouse, those would be 0.8 and 3.2 mg/kg, I believe). Intranasal dosing (long known as a peptide administration route, b.i.d. for two days, showed effects in the mouse cells consistent with GSK3 inhibition. I’d have liked to have seen more on systemic effects, but this is a good start.
There’s also some more direct pharmacokinetic data in the paper, though. Encouragingly, L807myr is stable for at least two days in mouse serum, which is already something that a lot of peptides wouldn’t pass. Its half-life after i.v. administration is 5 to 6 hours, which is better than I’d have expected. The authors are particularly interested in GSK3 as an Alzheimer’s therapy, and were able to show that the compound crosses the blood-brain barrier even after such an i.v. dose, which does surprise me. I can certainly believe it after intranasal delivery, but it’s interesting that it makes it in by a more traditional dose as well. A single i.v. dose at 24 mg/kg did not seem to lead to any overt toxicity, which is encouraging (but still nowhere near an actual tox workup by any drug development standard).
In keeping with the Alzheimer’s focus, the paper also shows behavioral improvements in a mouse model (5XFAD). I’m not going to make nearly as much of that as the paper does, because I just don’t believe mouse models of Alzheimer’s. But on a cellular level, they also show reductions in beta-amyloid and amyloid fibrils in the treated mice, which is what people have been looking to see in GSK3 inhibitors in this disease.
So overall, this is an interesting paper, particularly from a mechanistic standpoint. But this work has a long way to go. Other GSK3 inhibitors have also shown activity in rodent models of Alzheimer’s, for what that’s worth, but have been undone by (among other things) toxicity in the normal control animals. It’s very much worth noting that this new work shows that mouse-model activity taking place over 120 days of treatment, but there were no normal control animals, just 5XFAD mice with and without dosing of L807myr. But that last reference mentions that “in control animals, modest GSK3 inhibition itself induced neurodegenerative markers, inflammation and behavioral deficits“, which is not what you’re looking for. This new paper presents a nice-looking GSK3 inhibitor, but some of the problems that other compounds in this area have run into have been because they’re nice-looking GSK3 inhibitors. It is an important enzyme, as emphasized before, and you don’t get to choose what happens when you inhibit it. In other words, an interesting new method to make a really selective inhibitor may well not be enough to get you anywhere in the real world. We’ll see!