Skip to Content

Chemical Biology

Come One, Come All to These Kinases

Why do some proteins in a family prove very hard to target, while others bind a whole list of inhibitors? This paper takes a look at a particularly dramatic example in the kinase field. That’s a good place for studying such things, since there are a lot of kinases out there, and a lot of kinase inhibitors. Many of those compounds have been profiled across some part of the “kinome” to look at their selectivity – and believe me, some of them (even or maybe even especially approved drugs) are blunderbusses. But this analysis flips the question around, building on this 2017 work towards establishing a standardized data set for kinase inhibition, and asks about the apparent specificities of the enzymes themselves for small-molecule inhibition.

In general, across 398 enzymes you find a mean of about 17 inhibitors for each one in the data set. At one end of the scale, there are kinases out there that have only one known inhibitor. This might partly be due to lack of clinical interest, but remember, this is the result of taking all sorts of known kinase inhibitors and running them across enzymes that they were never targeted for, so that flattens out the bias to a good degree. And at the other end, there is a cluster of eight kinase enzymes that bind a mean of 99 of the known inhibitors (!), across a whole range of structures. Some of these in fact (like YSK4 and DDR1) have not been the targets of many drug discovery efforts, which helps rule out discovery bias as well.

What’s different about them? Phylogenetically, the eight enzymes are not particularly related in any evolutionary sense. X-ray crystal structures provided what looks like the answer, though. As kinase fans know, there are two major binding modes for such compounds, “DFG-in” and “DFG-out”, referring to a particular conserved stretch of Asp-Phe-Gly. There are successful kinase inhibitors in both binding category; the DFG-in conformation is the catalytically active one under normal conditions, and has traditionally been considered the more stable one. What’s odd is that these promiscuous enzymes, when exposed to classic DFG-in binding inhibitors, bind them in the DFG-out mode instead. It appears that these proteins have a particular salt bridge elsewhere in their structure that stabilizes the DFG-out conformation, and this opens them up to a huge range of potential binders. (That might well also explain the low catalytic activity displayed by many of the group, too).


The comparison between the classic “kinome tree”, which is based on sequence homology and the differences in inhibitor promiscuity are interesting. Anyone who’s worked in the field will recognize the shape of the tree at right; it’s been with us since the canonical 2002 kinome paper. But you can see from the overlay of this current work that the salt-bridged DFG-out enzymes (large green circles) are scattered through the tree in ways that the standard sequence analysis would not have made clear. This particular structural motif would appear to have emerged more than once – you should never assume that evolutionary history, as revealed by sequence, is going to tell you a straightforward flowing narrative.

Why some of these kinases have adopted what seems to be a deliberately less active lifestyle, with the DFG-out form stabilized, is open for speculation. The way that this conformation opens them up to a wider range of small-molecule inhibitors is probably a side effect; kinases aren’t regulated in the cell through endogenous small molecules that bind to the active site in this way. And it’s important to remember that this doesn’t mean that you can’t find a selective inhibitors of (say) DDR1. You most certainly can. What this work tells you, though, is that finding starting points for such chemical matter is going to be relatively easy.

20 comments on “Come One, Come All to These Kinases”

  1. Hap says:

    The less-selective kinases seem clustered on one or two branches of the kinase tree, though (and primarily in a smaller sub-branch at the top). Is there anything that those kinases do that makes them more amenable to flexibility?

  2. sgcox says:

    The kinase inhibitor set is not a some unbiased collection of molecules – all have been made to inhibit a specific kinase in mind. Most of the primary oncology targets are receptor tyrosine kinases which is exactly the branch with most green circles. What we see here is likely the bias of our research focus, not the inherent amenability of kinases for inhibitors.

    1. Magrinho says:

      Maybe… There is clearly bias in terms of the targets we choose. P38 anyone?

      There is bias at many levels including the kinds of compounds that reside in screening collections.

      But if you look at HTS data (and follow up re-synth/test) from LOTS of kinase screens, you will probably conclude that some kinases are harder than others to target with small molecules.

      1. sgcox says:

        That is absolutely true and will be nice to see some analysis published.
        One feature of “plastic” and easy targeted kinases is that they are up in signalling tree.
        TRK of course, but also YSK1 and MEK5 are on top of signal transduction.
        Might be a mechanism of a tight control.

  3. Daniel Barkalow says:

    I’d be interested in how other alleles of the genes for these enzymes behave. I could imagine the motif being something which offers a way to have variants with the same effect but significantly different amounts of activity, and the cluster of 8 is really all the kinases from a quarter of the kinome where the wild-type genetic knob is set to simmer. I wouldn’t really expect a shared genetic lineage for all the kinases that don’t do too much of whatever it is they do. (It would also be interesting to see if there are variants of enzymes with similar sequences that have much less activity that the wild type and are susceptible to inhibition by lots of things; very-effective DDR1 might just not exist because it’s lethal.)

  4. John Adams says:

    Has anyone figured out how to find out how a specific kinase is/is not inhibited by the specific molecules tested? If so, please share the details…

    1. Barry says:

      overwhelmingly, the kinase inhibitors are ATP-competitive (ParkeDavis’ MEK blocker notwithstanding). The block kinase activity by blocking the binding of the ATP

  5. Curious Wavefunction says:

    Neat study. Never ceases to amaze me how evolution has used the most specific and subtle of molecular interactions (like salt bridges) to modulate biology across the board. The great tinkerer indeed.

  6. An Old Chemist says:

    @Curious Wavefunction: “The great tinkerers are the aliens at a far away galaxy who started life on earth, and who are likely still tinkering with us. The co-discoverer of the structure of DNA, Francis Crick’s book “What a Mad Pursuit” tells that at a meeting of biologists, he and many of his fellow scientists agreed upon this hypothesis that life on earth is a guided evolution. Crick writes that these aliens have not yet visited us because we are still at a primitive stage. Who else can design >550 kinases with just subtle differences from each other in the structures, who do amazing biology whil we are busy doing our day-to-day things or even asleep. THe automated machines embedded in our bodies tiny cells certainly are better than Bruce Merrifield’s solid-phase peptide synthesis machine! Evolution certainly could not have perfected itself to have arrived at such sophisticated work.

  7. cmstop says:

    that was awesoem

  8. Cb says:

    I remember the times (~20 years ago) that medicinal chemists thought that it was impossible to make kinase inhibitors with sufficient selectivity: all kinase domains showed similar structural biology, used ATP in a similar conserved pocket and indeed the complex molecule staurosporine inhibits the kinome broadly. So you better stopped your research on kinases…..

    Nevertheless, despite this view some stubborn medicinal chemists (e.g. Zimmerman at Novartis: Imatinib) did create kinase inhibitors with desired selectivity, safety and efficacy. Only many years after the introduction of the first kinase inhibitor imatinib (Gleevec, 2001) we know that it stabilizes the inactive DFG-out state and not the active DFG-in state. Furthermore, we are not surprised any longer that every year more ‘selective’ kinase inhibitors are approved…even irreversible kinase inhibitors with a good selectivity profile (e.g. acalabrutinib).

    Imagine the inactive DFG-out state was already known 20 years ago and med chemists had to convince management to start a program to discover compounds that bind to the ‘inactive state of a kinase’; what about imatinib discovery and how many kinase inhibitors would have been arrived yet.

    Perhaps med chemists may see more opportunities for kinase inhibitors if these DFG-in/out states (including their crystal structures) are approached with a less static view and the regulatory control of individual kinases is taken into account. For instance, it would appear that AurA ( in addition to the inhibitory DFG-out state) employs an autoinhibited DFG-in substate for additional regulatory control. Other families are also regulated by autoinhibited DFG-in states, including the AGC kinase MSK1, CDK and Src-family kinases.

    In a recent review (Biochemical Journal (2018) 475, 2025) about the multifaceted allosteric regulation of AurA you get some nice hints for new opportunities. For example, some hypothesis that DFG-in inhibitors would show different effects from DFG-out inhibitors:

    “As Tpx2 enforces the DFG-In state, it might be predicted that the spindle-associated pool of AurA would be less susceptible to DFG-Out inhibitors, and more susceptible to DFG-In inhibitors. Conversely, since the phosphorylated enzyme found at the centrosome appears to still substantially sample the DFG-Out state, it is possible that this pool of AurA can be selectively targeted with DFG-Out inhibitors.

  9. barry says:

    We learned (after several companies had launched and abandoned Raf inhibitor programs) that depending on the conformation of B-Raf that your inhibitor stabilized, you could either block or–paradoxically–enhance signaling down the ras-raf axis. Because one conformation of B-Raf–even if it were phosphorylation incompetent–activated C-Raf.

  10. hn says:

    It will be interesting to see whether druggability can be extended to other enzymes with ATP-binding.

  11. A young chemist says:

    @An Old Chemist

    Just curious, but who engineered our alien mother’s kinome?

    1. Dr D says:

      The flying spaghetti monster? Lizard people? The Welsh?

    2. An Old Chemist says:

      @A young chemist, you better address your question to Francis Crick, he is still sharp as ever and has been creating controversies. Just a few days ago, he was stripped out of his honors for passing a racial remark on a radio talk show. BTW, our universe is 11.5 billion years old and our sun is only 5 billion years old. Therefore, there must be suns in universe, which planets like ours, where life evolved 5-to-6 billion years earlier. Thus, aliens living there, are ahead of us humans by a big margin. On our planet earth, most of the scientific/technological progress has been made only in the last one-and-a-half century. So, these aliens have likely figured out everything that there is to be figured out, including engineering us and our kinases. Then, Einstein once said “I believe in the God that put the universe in motion, but not in the God who sits on the clouds and looks after us on a day-to-day basis.” ……So, these two genius scientists have provided answer to your query.

      1. Russ says:

        That was co-discoverer James Watson; Francis Crick passed in2004.

  12. petros says:

    Magrinho mentions p38 but there were a stack of inhibitors of this kinase, mostly from SB, before the target was identified as a kinase. They even registered the term CSAID to describe such inhibitors.

  13. kinasebod says:

    I just skimmed this paper but will take a lot of convincing that this is a meaningful result. We solved multiple co-crystal structures of one particular kinase I won’t name with different inhibitors and under different conditions, and we found the DFG motif ended up in subtly different conformations almost every time.

    That’s not really so surprising since in may kinases this motif helps to control the activation state, so of course it needs to be flexible. When we solve a crystal structure we take the domain out of context so whether this represents the active state or not is a lottery.

    Seems to me that they have seen an=1 observation for DDR1, which they have generalized way beyond what the data supports.

  14. hse says:

    FLT3 is not mentioned in the paper, and its class III receptor tyrosine kinase cousins — PDGFRa, c-KIT, CSF1R — are highlighted as quite promiscuous. In another paper that I can’t recall, Flt3 was found to be the most promiscuous kinase of all among a large panel. Any insight on the omission of FLT3?

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.