Skip to main content

Chemical News

CDK Inhibitors: Purchase With Caution

Cyclin-dependent kinase (CDKs) have been drug targets for quite a while now. There are 20 different ones, and they help to regulate a whole list of important functions, particularly involving the cell cycle (which has made them of great interest in oncology research). There are three approved drugs in the area so far: Kisqali (ribociclib), Ibrance (palbociclib), and Verzenio (abemaciclib). Here’s a 2014 post on that second one and its history. All of these are CDK4/6 inhibitors, approved (thus far) as part of treatment regimes for breast cancer. These CDK subtypes  are involved in the “checkpoint” after the G1 phase of the cell cycle (which requires phosphorylation of the Rb protein), and inhibiting them keeps this from taking place and interrupts cell division.

That’s the CDK area that we know the most about, and have the most tools to address (CDK1 and CDK2 also have a good amount of work done on them). They’re a pretty complex group, though, and many of their functions are still being worked out. The usual complications are there: the same enzyme subtype can be performing several quite different functions on different substrates in the cell, for example. Alternatively, sometimes you find things that look different at first but can be traced back to new functions of the same substrate protein. The CDK family is evolutionarily rather conserved (you can swap the human forms into yeast and they seem to do just fine, for example), so they’re not something to be messed around with lightly.

But about those tools. . .here’s a new paper that takes a look at the reported compounds, and finds some confusion. The authors (a Czech team) note that there are around 100 inhibitor compounds out there with reported activity, but (as is so often the case) taking those reports at face value is not recommended. They’ve evaluated a list of compounds that needed to be better characterized against CDKs 1, 2, 4 ,5, 7, and 9 in enzyme assays, and in cell assays as well.

As it turns out, there are a lot of mischaracterized compounds out there. CGP74514A, for example, is usually described as a CDK1 inhibitor, but it’s more potent against CDK2 (and shows activity against plenty of other kinases besides). Benfluorene and elbfluorene are available from several prominent vendors as selective CDK1 inhibitors, too, but this group was unable to confirm any CDK activity against any subtype, and they don’t seem to do much of anything to HCT116 cells, either (!) The most suitable CDK1 compound, if you’re in need of one, appears to be RO3306.

The pattern continues on the other subtype compounds. NU6140 is sold cheerfully by many vendors as a CDK2 inhibitor, but this work shows that it’s more properly described as an Aurora kinase inhibitor, which will make your CDK2-based conclusions hard to stand by. And commercially available CVT313, which the catalogs say is a “selective and potent” CDK2 compound, is actually a dual targeting CDK2/5 inhibitor. And you can buy GW8510 as a CDK2 inhibitor as well, but you’d better not: it’s actually a CDK 1/2/5 compound, hits other things besides, and if you’d looked at you could know these things before getting tangled up in trouble. Sigma-Aldrich might want to consider updating its catalog. In the end, the authors find that none of the commercial “selective CDK2 inhibitors” are suitable for cellular experiments.

Now, if you want to inhibit the CDK4/6 types, you’re in luck. As mentioned above, these have marketed drugs against them, and the companies involved have put in the work to show that they do what it says on the label. You can buy other compounds in this area, but it’s not a good idea: NSC625987, for example, is reported as a CDK4 inhibitor, but the Czech group report that they can’t get it to show any activity at all (up to 20 micromolar) against any of their CDK subtypes, and they note that this lack of activity has been reported before. But that’s not slowing down Tocris, EMDMillipore, Sigma-Aldrich, Santa Cruz, Axon, Fisher, Apex, Probechem and several others. They all have it listed as a selective CDK4 compound, and let the buyer beware.

The paper has plenty of further information, but the authors have compiled their results into a publicly accessible database to help out researchers in the field. If you are doing CDK-dependent assay work, you should absolutely have a look – there might be some unpleasant surprises waiting for you, but the time to find out is now. Overall, the lessons are clear: just because that’s what the vendor says it does, doesn’t make it so. Just because the literature says that’s what a compound does, doesn’t make it so. And just because others have used this as a control compound doesn’t necessarily mean that you should, too. Check it out.

You may think that “due diligence” is a phrase from the business world, but that is far from true. Every time you pick up on an assay, a probe compound, a cell line, what have you from the literature or a catalog, it is incumbent on you to do the homework and make sure that you’re using what you think you’re using. Make sure it’s what it says on the label. Then even if it is, make sure to the best of your ability that it’s going to do what you brought it in to do. Otherwise, you are inviting the laziest, sloppiest links in the whole chain to do their worst to you and your research. Why would you allow that to happen?

28 comments on “CDK Inhibitors: Purchase With Caution”

  1. Peter S. Shenkin says:

    “For due diligence, you have to do diligence.”

  2. Isidore says:

    I would imagine that vendors rely on the published literature and do not carry out their own testing to confirm reported activity. I mean supermarkets sell tomatoes, zucchini, onions, peppers but they have not necessarily made sure that these ingredients will work well in ratatouille, nor would I expect them to.

    1. Derek Lowe says:

      True. But in this case, we have limes being sold as oranges, tomatoes being sold as plums, and yellow-colored balls of styrofoam being marketed as lemons. . .

      1. Isidore says:

        Fair enough. I also have bought, over the years, stuff from vendors that turned out to be not as pure as stated on the label (and in the catalog) and a few times it was completely the wrong thing. In cases like these complaints are, of course, warranted. But if a peer-reviewed paper claims that compound A has activity B and a vendor copies this claim with the appropriate literature reference under the listing of compound A in their catalog I don’t see how they are to blame if the referenced work was wrong. I agree that Sigma and everybody else should make necessary updates to their catalogs when information like that in the paper from the Czech group becomes known.

    2. Hap says:

      If they’re selling funny-shaped metal things as components of the transmission for a Ford Focus, though, then you’d actually expect them to fit in the transmission for a Ford Focus. If they don’t, they don’t have any other reasonable use.

    3. kindom says:

      SD-282, a p38 inhibitor, was misleadingly described in its article as “p38alpha-selective mitogen-activated protein kinase (MAPK) inhibitor, indole-5-carboxamide”. This led at least one vendor to market indole-5-carboxamide itself as a selective p38 inhibitor.

      I wrote to the authors and to the vendors to point out that they should have said it was an indole-5-carboxamide CONTAINING inhibitor. The vendor was great, they immediately pulled the product from their website. The authors just ignored me.

  3. JB says:

    This issue is not limited to the CDKs. Kinase inhibitors are selective – as long as you test the “right” kinases. Most inhibitors are truly only tested against a very small subset of kinases. This is a general problem with a lot of inhibitors commercially available, kinases as well as other targets. If you don’t test it yourself, you will often get some unexpected surprises. And as aside, a lot of commercial vendor websites may not get updated when data like this paper comes out.

  4. MarkySparky says:

    I seem to remember reading either Derek or a comment years ago, paraphrasing: “biologists believe the label” (structure/purity/etc) and “chemists believe the pathway diagram” (in vivo relevance). Has served me well through grad school and industry….

  5. Curious Wavefunction says:

    Wow, turns out that a CDK7 inhibitor I had designed in grad school makes the cut in their study (although it does indicate a bit more promiscuity than what we had seen).

  6. Peter S. Shenkin says:

    It is well known that assay data published in the literature is very often faulty, or at least that very often, multiple assays of the same substance performed by different groups, and especially using different methods, are very often in disagreement beyond normally expected error bounds.

    Thus, if you really care about how your flashy new drug candidate stacks up against the competition, you need either to repeat all the assays yourself or else have someone else whom you have good reason to trust do it, employing, for all the assays, the same methodology in the same lab. Then at least you can hope that the relative results are meaningful, even though the absolute results can still not necessarily be trusted.

    In the current situation, there may be compounds whose results for, say, binding against the desired target were accurate, but for which not too many (if any) of the other kinases were in fact tested. Thus, the activity may be right but the selectivity only partially, if at all, tested. I’ve not gone and read the article and its references to see how many ligands this is the case for, but either way, articles like this are very important, especially for modelers, who rarely lack the resources (and, dare I say, the motivation) to be certain that the data that they are trying to reproduce are in fact themselves reproducible.

  7. ScientistSailor says:

    *If you are doing CDK-dependent assay work, you should absolutely have a look – there might be some unpleasant surprises waiting for you, but the time to find out is now*

    But, but, but, that makes life hard and I just want to publish a paper so I can renew my grant…

  8. tlp says:

    For NU6140, a quick ChEMBL query will give similar results the authors of the paper above got (good Aurora inhibition with poor Cdk2 activity). Funny thing is that the study referenced in ChEMBL are from 2013 ( and done by Millipore, which still lists the molecule as selective Cdk2 inhibitor on their website. Go figure…

  9. Thoryke says:

    I’m reminded of all the cell lines that had been contaminated with HeLa cells [Henrietta’s revenge?]

  10. Ron says:

    Is there really any such thing as a truly selective ATP competitive kinase inhibitor? (And 10 fold doesn’t count.) I actually never worked in the kinase area but I was always skeptical. I thought that their panel wasn’t large enough or they hadn’t discovered the other kinase that it was hitting. (But I did laugh out loud when some group published a paper many years ago lauding their multi-kinase inhibitors……as if that was a feature that other groups could not easily reach.)

    1. Barry says:

      Chiron published a very selective GSK-3 inhibitor. Exactly why it is as selective as it is (and why the SAR is as steep as it is) isn’t clear to me.

    2. ScientistSailor says:

      There are some very selective PI3K and mTOR ATP-competitive inhibitors…

      1. Barry says:

        To the best of my knowledge, no one has a small molecule that is very selective among PI3k alpha, beta, and gamma. But yes, the ATP binding site of the PI3ks is significantly different from that of protein kinases.

        1. Researchfella says:

          And yet, a compound called ASN003 is somehow highly selective for both PI3K and RAF kinases.

  11. sgcox says:

    I think some ATP competitive p38 inhibitors, like vx-745 are extremely selective

    1. Ron says:

      Well VX-745 inhibits MKK6 which phosphorylates more than just p38. And it is only 20 fold selective for p38 alpha over gamma. And the one report I just read looked at ~25 kinases. And it tanked for tox in animals in Phase II. So, what else was it hitting? Was it mechanism based tox or perhaps other kinases? The animal tox was neurologic which was not the usual tox you see with the TNF convertase inhibitors. Granted TNF antibodies carry neurologic black box warnings for increased incidence of developing MS and TNF antibodies actually increased disease activity in MS patients, paradoxically. Maybe you are right but I haven’t been convinced yet that kinase inhibitors are selective and I think that’s the reason why they all seem to be used for cancer where anything goes, so to speak.

      I used to work for a company that had a project targeting a kinase inhibitor for obesity. Luckily, I didn’t get put on that project because I thought it was a pretty stupid idea. (It died a slow painful death.)

      1. Hap says:

        Tox for animals in Phase II? Isn’t that sort of late to be testing in animals?

        1. Dionysius Rex says:

          Could have meant something like reproductive tox in rabbits?

    2. kindom says:

      The 2 most selective ATP-site kinase inhibitors I know are lapatinib (EGFR/ERBB2) and losmapimod (p38a/b) which each hit just 2 closely related kinases in panels of 300-400 or so. The Pfizer MEK1 inhibitors are completely selective as far as I know but they bind in a unique site just off the main ATP pocket.

  12. kriggy says:

    Wow I would never thought that I will see a work from my institution mentioned here and even less from guys that I know.
    Its not my field but it looks like a big deal makes me wonder, how many similar cases exist? Im not talking about CDK but about other molecular targets ?

  13. Frank L says:

    check out this “specific” JNK kinase inhibitor, with a working concentration range 10-50 uM(!)
    Now do a Pubmed search with this compound, you will find >3600 papers using this compound…

    1. Barry says:

      yes, that’s a junk inhibitor

  14. beentheredonethat says:

    Take the well-validated CDK/4/6 inhibitors, approved drugs with proven selectivity, run the assay under the wrong conditions (with for e.g.slightly different substrates) , you are in for a big surprise too. We see this all the time!

    Admit that I am yet read the new paper, but what makes us think that the assays in the new paper was run under physiological conditions as much as possible??

  15. I totally agree and this happens with a lot of inhibitors sold by a lot of compound vendors!

    One example that I always show in presentations because it worries me a lot is iniparib, which is still sold as a PARP inhibitor by many vendors despite it was clearly shown to be a non-specific cysteine modifier after it failed in the clinic. Like here:

    It is challenging to address this situation. Recommending to do a due diligence is of course a very good idea, particularly checking that the compound is what it says on the label! But I also think is beyond many lab’s capacity to thoroughly check all the chemical probes’ key properties like the selectivity of the chemical probes they need at scale. I think we must leverage current knowledge to point researchers to the right direction when it comes to chemical probe selection and alert them on the existing limitations.

    The Chemical Probes Portal ( are doing a nice job at pointing researchers to the right chemical probes but of course it is challenging that expert reviewers can review and assess all the potential probes for all the targets and review their assessments every time new data is published, like in this case of CDKs.

    We have recently developed a complimentary approach. Probe Miner ( [Antolin AA, Cell Chem Bio, 2018] uses large-scale data from public databases and statistical analysis to objectively assess probes for any given target. The idea is to give researchers user-friendly access to available knowlege and alert them on gaps within it so they can make an informed choice. We find nice correlation with the Portal and of course wider coverage but there are also some limitations, particularly with the available data.

    As illustrated in this post for CDK inhibitors, the key selectivity data had not been reported for many compounds, making any objective assessment incredibly challenging. We also find that many relevant chemical probes are published in journals that fall outside the normal journals indexed in public med chem databases. Therefore, public databases are missing very relevant data, particularly on the selectivity of these potential chemical probes! We are starting to curate and integrate this missing data but it is not an straightforward job.

    Beyond technical improvements, ensuring that a broader target profiling is done and reported would be very important as well as getting the message across regarding probe selectivity and the importance of selecting chemical probes appropriately, like in the recent Blagg & Workman perspective published in Cancer Cell trying to reach out to cancer biologists:

    Overall, a lot of work to do if we want to address this challenging situation but hopefully the situation is improving!

Comments are closed.