Skip to main content


Why Did This IDO Inhibitor Trial Fail?

Let’s put this one under the “This is why you run clinical trials” heading. A number of companies have looked at inhibitors of indoleamine 2,3 dioxygenase (IDO) over the last few years. The enzyme turns tryptophan into kynurenine, a pathway which (among other things) regulates immune function. Evidence has accumulated that many tumors have increased IDO activity as part of their ability to decrease T-cell response and evade immune attack, partly through the decreased tryptophan levels in the cells. (They’re probably also doing this through a related kynurenine-producing enzyme, tryptophan 2,3-dioxygenase (TDO), which was earlier thought to be found in only liver and neuronal tissues).

What with immuno-oncology being as blazing hot a research field as it is, a separate mechanism that could be combined with the current therapies is definitely going to get a lot of attention. In 2015, Genentech and NewLink Genetics reported that a NewLink IDO inhibitor (GDC-0919, navoximod) combined with anti-PD-L1 antibodies gave robust increases in activity in preclinical tumor models:

The combination of IDO1 inhibition with PD-L1 blockade has strong rationale since both checkpoints are co-expressed in many cancers and inhibit immune responses via complementary mechanisms. Using mouse syngeneic tumor models we present further evidence of the therapeutic potential of combining IDO1 inhibition with anti-PD-L1 blockade. We observe improved depth and duration of responses in mice treated with the combination regimen versus anti-PD-L1 blockade alone.

This combination was already in the clinic by the time the paper appeared, with a fair amount of optimism from observers. But we now have a readout on how it’s going, and the answer is “not well”. Genentech is handing back rights to GDC-0919, after data presented at the recent ASCO meeting showed only a 10% response rate across several tumor types. You can see from that link that some analysts were trying to remain optimistic, but Genentech apparently doesn’t share that glass-half-full outlook.

Something’s wrong, but what? Incyte is developing an IDO inhibitor of their own (epacadostat) and combining it with Merck’s Keytruda is providing the results that one would hope for. So the mechanism is likely to be sound. Complicating the picture, though, is that the Roche/Genentech PD-L1 antibody used in this study (Tencentriq, atezolizumab) had a recent unexpected failure in bladder cancer. It will be tempting for NewLink (and their investors) to think that the antibody was the problem, but there’s just not enough evidence to make that case one way or another. NewLink itself is under a lot of pressure after they’ve had disappointing results with their other IDO inhibitor, indiximod, in breast cancer, and it might be tough to round up another partner at this point. There’s a fair chance that we may never find out why GDC-0919 did what it did, but I hope that we eventually get some idea.

11 comments on “Why Did This IDO Inhibitor Trial Fail?”

  1. Chrispy says:

    The scientists at BMS can’t be too pleased with the purchase of Flexus for $1.25 billion. Derek wrote about that here:

    Now that they’ve gutted internal research, Big Pharma is desperate for the internal candidates they can’t generate themselves. Unfortunately for all of us, these are purchased with so little understanding that their price tag is way out of line. They’ve just decided to buy what everyone else is buying instead of doing research.

  2. Daniel Barkalow says:

    It’s this sort of thing that makes me wish that failed clinical studies produced a lot of basic science data. Does GDC-0919 just not inhibit IDO in humans? Does it not get into these cells? Is there an alternative to IDO in the same immune inhibition mechanism? Is there a third immune inhibition mechanism? It seems like the proteome of tumor cells in patients that didn’t respond to GDC-0919 and Tencentriq would be really useful information for everyone else. It seems like we might be missing the opportunity to study some very interesting cells that were created accidentally and which wouldn’t really be ethical to recreate later, and it would be nice if NewLink could recoup some of the costs of this trial by selling Incyte (et al) information that might save them money later.

    Occasionally, you see studies which show that the drug did what it was designed to do (and still didn’t help), but so often all we get out of it is the information that the drug company isn’t going to be able to sell the drug.

    1. bhip says:

      I have no personal insights on GDC-0919 as a drug but sometimes a clinical trial simply reflects on the drug (or trial design) not the mechanism (i.e. Incyte IDO cmpnd seemingly helping Merck anti-PD-1). Re: the drug, what is its protein binding, tumor distribution, whole cell potency, Cmin., etc. Re: trial design, not all tumors express IDO1 to any great extent- was that checked or did they simply BMS it & take on all comers?

    2. tangent says:

      I wonder if NIH could chip in money to augment clinical trials for an appendix of public data that’s useful for basic research more than for pharma. It would be tricky to prevent exploiting the taxpayer given that the company holds so many of the cards — also there are limits before you’d be exploiting the subjects — but it could be opportunities that just don’t happen otherwise.

      A particularly tricky line is when the question is whether the mechanism is any good (effective + not inherently toxic), and the company has only limited interest in that — say they only have one candidate left, and if it doesn’t work they don’t think it’s worth starting another. But other companies may be wasting money and patients in running trials on their candidates. Maybe a 2x-cost supertrial could rip the heart out of the mechanism and we could collectively avoid running 5 trials to completion. But how do you line up all the incentives…

      Possibly the private approach would work as well as anything, that the first company invests to generate data they don’t need themselves but try to sell to a competitor. My gut says nobody would buy it because “Not Researched Here” — would the leadership chain be willing to kill the company’s effort because somebody else’s research says it’s futile? And maybe that vitiates the whole idea whoever funds it.

    3. Anon says:

      “It’s this sort of thing that makes me wish that failed clinical studies produced a lot of basic science data.”

      Usually you’re lucky if they answer the one question you were asking – without raising many more!

  3. checkpoint guy says:

    The NewLink compound is not only a weak inhibitor of IDO, it also blocks TDO. By contrast, the Flexus/BMS compound BMS-986205 potently inhibits IDO while sparing TDO (epacadostat also spares TDO, but is significantly weaker against IDO than 986205 acting via a different mechanism).

    Add the additional uncertainty of PD-L1 blockade vs PD-1 and interpretation of Tencentriq-GDC-0919 clinical failure becomes even more convoluted.

  4. Anon says:

    So is the glass 10% full or 90% empty?

  5. Alchemist says:

    I don’t know if it was a question of a symbol gone awry, but in their last ASCO presentation, the IC50 in the biologically relevant in vitro assay (the ability of Navoximod to restore T cell function) was given as 0.1 mM. So unless it was really 0.1 uM and the symbol messed up, the compound is not really potent in a cellular functional model. And it certainly does not reach 0.1 mM concentrztions in patients’ plasma.

  6. jc says:

    bad targets?
    bad compounds?
    bad trials?
    bad combinations?
    In this case it is likely because of all above 4!
    The target is not well validated. The compound is not potent. Its PK is terrible. PD effect was not shown in any tested species (mouse, rat, dog, and human). Why should such a compound work with anti-PD-L1?

  7. Brian says:

    According to literature survey, most of the studies revealing IDO1 is an immune modulator are based on “IDO1 gene KO study”. Although lots of groups have developed very potent IDO1 inhibitor, the “inhibition of IDO1 enzymatic activity” by chemical inhibitor only showed moderate cellular immuno-response. What if IDO1 has other biological function rather than tryptophan metabolism? If it is the case, IDO1 inhibitor may not show any clinical benefit. As far as I know, IDO1 also involves in signal transduction and can produce significant ROS, which may also regulate the activity of immune cell (such as Macrophage) and tumor micro-environment.

  8. Barry says:

    That a KO is viable is encouraging–it says that blocking a target’s activity won’t be lethal. But because a single protein (and drug targets are overwhelmingly proteins) often has several roles (catalysis, scaffolding…), a KO is a treacherous surrogate for e.g. an enzyme inhibitor.

Comments are closed.