Skip to main content


PI3K Inhibitors: You’re Doing It Wrong

Now here’s an interesting connection between cancer and metabolism, with what look to be direct implications for therapy. A large research team (mostly working out of Weill Cornell) reports some new and important details about PI3K inhibitors, a class of kinase inhibitors that has seen a very large amount of development work indeed. I’ve worked on such a project myself, and there will be plenty of readers who have as well. The phosphoinositide-3-kinases are a whole family of enzymes, and getting selective inhibitors is (mostly) possible, albeit tricky. And since phosphitidylinositols of various sorts are absolutely ubiquitous signaling molecules, these enzymes sit along a ridiculous number of cell biology pathways.

So there’s a lot to figure out in this area, but one thing that’s been reasonably clear is the application to oncology. There’s an insulin-activated subtype knows as p110-alpha that has been known for a long time to be frequent mutation target in many cancer types (for example, it’s the most frequently mutated gene in glioblastomas, which should be enough to tell you right there that it’s bad news). These mutations are generally activating ones leading to inappropriate kinase activity whether there’s insulin signaling telling it to get moving or not, so finding inhibitors has been an obvious drug discovery strategy.

Note that “obvious” does not mean “easy”. People have been working in this field for many years now, generally trying to get compounds that are selective enough, and that also have reasonable pharmacokinetics once you’re closing in on that goal. If you want to go after glioblastoma, for example, you’ll need to get past the blood-brain barrier, which can be a constraint on some of the compound classes that have been investigated. But even going after tumors out in the rest of the body has been a challenge. If you go through the (voluminous) literature in this field, you find all sorts of worrisome article titles like “Challenges in the Clinical Development of PI3K Inhibitors” and “PI3K Inhibitors: Lessons Learned From Early Clinical Trials”, and these things are like five and six years old already. Ideally, you want to be able to match up the patients with the drug for maximum effect, but that has proven extremely difficult in this area, with a lot of trials showing no particular correlations between the type of tumor, what was known about its mutational state, and the response to any given inhibitor. “Who knows, maybe it’ll do something” is not a rationale that you want to take to the FDA.

Another problem with inhibiting p110-alpha is that it’s so crucial in mediating the cellular response to insulin. Disrupting that tends to lead to hyperglycemia, as you’d figure – it’s like the insulin wasn’t even there – but this generally resolves itself within a few hours. But what if the patient was insulin-resistant from the start (Type II diabetes?) Knocking out whatever remains of the insulin-signaling axis is the last thing those patients need, and this is in fact a major problem in the clinic. This new paper takes a closer look at this, and the team finds, rather surprisingly, that in some tumor types insulin is such a potent stimulator of p-110-alpha that this effect overcomes the effect of small-molecule inhibitors.

You wouldn’t have guessed that this is the case if you were just looking at the hyperglycemia (hey, insulin signaling is clearly impaired, right?) It looks like the PI3K inhibitors are causing insulin-axis trouble in normal tissue, but the tumors are able, perversely, to handle it a lot better. The insulin spikes seen after inhibitor treatment in mice are actually correlated with increase glucose uptake in tumors – they’re still managing to respond – and with increased signaling in cellular growth pathways. The pancreas responds to the disruption of the signaling pathway, naturally enough, by secreting more insulin, so the system is in a state with both too much glucose and too much insulin floating around. This is the same thing you see in Type II diabetics – they’re often hyperinsulinemic, because the pancreas is frantically trying to get some sort of response out in the (insulin-resistant) peripheral tissues. (In Type II patients, this process can continue until the islet cells just wear out, in which case you’re in even bigger trouble). But this hyperinsulinemic state is actually just what the tumor cells need to deal with the PI3K inhibitors. No wonder the clinical responses have been mixed.

The paper goes on to try some methods to disrupt this signaling axis by using the same methods that you would use for Type II diabetes. Metformin unfortunately had no effect, but treatment with SGLT2 inhibitors improved things, as did placing the rodent model animals on a ketogenic diet. Important note: there’s a lot of craziness out there around “keto” diets, and this does not mean that a ketogenic diet prevents cancer or is a treatment for it. No, what it means is that if you’re a cancer patient taking a PI3K inhibitor, a ketogenic diet could help the drug to work, and that’s it. In fact, in at least one rodent model (AML), putting the animals on the diet actually accelerated the cancer before the PI3K inhibitor treatment began, so be aware.

This effect wasn’t seen in tumor-bearing mice whose insulin receptor signaling was knocked down by other means, and the diet effect could be largely reversed by administration of more insulin on top of it. Overall, it looks like the insulin receptor effect on tumor growth only really kicks in at quite high insulin levels. Which perversely is what the PI3K inhibitors are providing! The paper demonstrates this effect across a range of different tumor types and with a whole list of known PI3K inhibitors with a range of selectivities (BKM-120, BYL719 (alpelesib), GDC-0941 (pictilisib), GDC-0032 (taselisib), GDC-0980 (apitolisib), and Aliqopa (copanlisib), which is already on the market. It looks like quite a strong story to me. This should be tried out with investigational PI3K compounds in the clinic, and it really should have an immediate effect on current clinical practice: if a patient looks like a candidate for a PI3K inhibitor, there would seem to be good reason to consider either putting them on a ketogenic diet or administering an SGLT2 inhibitor. And at the same time, there would seem to be good reason, as the paper mentions, not to give them glucose-laden IV drips and nutritional supplements, since this would just exacerbate the insulin levels. A very worthwhile result!



6 comments on “PI3K Inhibitors: You’re Doing It Wrong”

  1. Barry says:

    Do we know that the blood-brain barrier is usually intact in glioblastoma? It has been argued that in this disorganized tissue, small molecule access shouldn’t be as hard as in wt brain.

    1. Druis says:

      Take a look at this paper “Is the blood–brain barrier really disrupted in all glioblastomas? A critical assessment of existing clinical data”. (Open access.) As with other brain tumours including metastases, it seems that there is a portion exposed to the blood and a portion behind the BBB which is hard to get at.

      1. Barry says:

        Oooph. Leaving part of a cancer is a lot like failing.

  2. Kthe Knight says:

    Could you add the paper?

    1. Jim Hartley says:

      I’m pretty sure Derek is referring to

      Lew Cantley, et al., in Nature, July/August 2018

  3. David Young MD says:

    Now Duvelisib has been FDA approved for lymphoma

Comments are closed.