Skip to Content

Chemical Biology

HAT Inhibitors: Interpret With Care

There are quite a few histone deacetylase inhibitors out there, from research tools to FDA-approved drugs. Those inhibit the enzymes that remove the acetyl epigenetic markers from histone proteins – but what about inhibitors of the enzymes that put them on? Those are histone acetyltransferases (HATs), and they’ve naturally been the subject of a lot of work as well. There’s a whole list of reported compounds (although nothing’s made it through the clinic yet), but how useful are they?

Not very, according to this new paper. The authors (a multicenter team from Brigham & Women’s Hospital, the University of Minnesota, the SGC, NCI, Monash, and Nanjing) investigate 23 reported HAT inhibitors, and right off the bat find that 15 of them are active in the ALARM-NMR assay. That’s a screen for thiol reactivity – you expose a solution of the La antigen protein to your compound of interest, and look for specific changes in its conformation when one or both of its sensitive Cys residues gets alkylated. The readout is “This compound can react with proteins”, and of course it’s up to you to decide what to do about that. If you’re using it as an in vitro tool compound, you’ll want to check the specific proteins you’re using – maybe things are still OK. But if you’re going into cells (or worse, in vivo), this throws suspicion on your phenotypic readouts, because there are so many other things that could be happening. You can do a whole proteome screen for reactivity, if you have the time, equipment, and money, but even that’s not going to pick up everything. In this case, the team went on to check reactivity with small-molecule biological thiols like GSH and CoA, and many of them formed adducts with these as well. (As a side note, the profile in these assays compared to the ALARM NMR data don’t completely overlap, in either direction, and the paper has a discussion of how to interpret such situations – if you’re dealing with potentially reactive screening hits, it’s definitely worth a read).

The data get worse. Those latter experiments were monitored by LC/MS, and that showed that six of the reported compounds are not even stable in buffer, but degrade to various by-products under standard assay conditions. Twelve out of the 23 also showed detergent sensitivity in a standard aggregation assay test as well. All of these problems are going to show up no matter what sort of assay readout you use, but it should also be noted that several compounds showed potential fluorescence problems as well (which is always something that has to be checked with any such assay; sometimes half of your “hits” are merely fluorescent interference compounds).

I’ve never worked in the HAT area myself, and I hadn’t paid that much attention to the reported chemical matter. Looking at the compounds, I’m not the least bit surprised that many of them are reactive or otherwise problematic, though: it’s a very unappealing collection (which even includes our old friend curcumin). How you’re supposed to make sense of cellular assays with such compounds is a mystery, and in fact, the paper shows that the great majority of reported HAT compounds cause nonspecific effects in cell lines. When your tool compound is indistinguishable from a stink bomb like rottlerin, you have problems: you may think that you’ve asked your compound to tell you all about some specific pathway, but in reality, you asked it to just go in and mess your cells up, any old way.

. . .our data show none of the reported HAT inhibitors contain a substantial number of attributes associated with high-quality chemical probes such as target potency and selectivity, tractable mechanism of target modulation, and meaningful target-based cellular activity12. Histone acetylation is a complex biological phenomenon, and we argue the use of suboptimal compounds to probe for HAT cellular functions leads to tenuous, and even false, scientific conclusions that cannot fully account for the contributions of off-target activity, promiscuous reactivity, and/or nonspecific cytotoxicity to cellular, biochemical, or phenotypic readouts.

. . .The cell-based data indicate nonspecific assay interference compounds can disrupt cell proliferation, decrease cellular histone acetylation levels, and even perturb HAT levels. These findings suggest that test compounds have the potential to be misinterpreted as useful HAT inhibitors if the appropriate mechanistic, selectivity, and analytical experiments are not performed.

Absolutely. Garbage in, garbage out is the law, and wishful thinking does nothing to abrogate it. As the paper shows, these compounds appear in hundreds of papers (and hundreds of review articles), so just reading the literature would give you the impression that there are plenty of useful HAT compounds out there, and that we’re figuring out what they do in cells. Not so. These results need to be regarded with skepticism until proven by other means. Looking at the chemical suppliers would give you the impression that you can order up useful compounds. Not so – the authors of this paper end up by being unable to recommend any of the 23 compounds they profile. There are, they note, HAT compounds being reported in industrial patents that look to be better, and further profiling of these by different labs may well validate them as useful tools. But what’s in the open literature now is not cutting it.

30 comments on “HAT Inhibitors: Interpret With Care”

  1. SP says:

    There’s a recent publication reporting a specific p300 inhibitor from a pharma company, they would probably test for that sort of off-target activity. I assume the SGC group didn’t test it given the timing, they were probably already in-press when the AbbVie paper came out. Includes biophysical characterization and crystal structure so I’d say it’s authentic- also just by eye you can see it doesn’t have the ugly reactive groups most of the reactive molecules do.

    1. Mike says:

      Looks like they did?

      “The future of HAT inhibitors is not completely bleak. Patent applications have been filed by industry related to cell-active p300-specific inhibitors with encouraging results, including low nanomolar potency in multiple biochemical assays, low nanomolar activity in multiple cell systems, and promising selectivity profiles62,63. However, a rigorous and peer reviewed evaluation of this science by other groups is needed for confirmation69.”

      Ref 69 is the paper you linked to.

      1. Mike says:

        Nevermind my comment, I see now that you said test rather than reference.

        1. SP says:

          Yeah, probably added in revision? And they didn’t have a ready source of the compound to test it.

          1. sgcox says:

            Abbvie kindly made it a donated probe

          2. copyeditor says:

            Looks like the Nature paper has the wrong structure for A-485, although that sulfur replacement of the carbon would be an interesting synthetic challenge.

  2. Peter S. Shenkin says:

    “The data get worse.”

    @derek, thanks for using data as a plural noun….

    1. Pennpenn says:

      I know it’s right, but it feels so wrong. Man I hate the English language…

    2. NotAChemist says:

      The singular of data is anecdote, right?

  3. Barry says:

    one typo:
    “rotterlin” should be “rottlerin”

    Plainly, covalent, irreversible inhibitors have some successes as drugs, but they have distinct challenges in testing and development

    1. Derek Lowe says:

      Freudian slip, no doubt – fixed!

  4. a. nonymaus says:

    So, previous workers in the field are all HAT, no cattle?

  5. Wavefunction says:

    False positives? That’s old hat.

  6. Dionysius Rex says:

    Sometimes you have to despair at the state of science – isothiazolones have been known for at least 15 years to be thiol reactive p300 inhibitors.

  7. MoMo says:

    As soon as we all can accept the fact that every drug is pleiotropic, with an average of 6 cell targets per drug, the better off this industry will be.

    Until then you are just fooling yourselves that you are making a specific drug.

    1. Barry says:

      To observe that the average drug hits six targets doesn’t preclude some of them from hitting one and one only.

      1. Wavefunction says:

        Yes, but the laws of physics (and especially weak VdW forces) guarantee that no drug will hit one and only one target. Even covalent drugs which were thought to be very specific for their targets have broad activity across the proteome. The question is how many other targets a drug hits and with what affinity, and whether that differential window preferring one target is good enough.

        1. Peter Kenny says:

          Paracelsus might have drawn our attention to the importance of drug concentration when discussing selectivity…

    2. Imaging guy says:

      I have to agree with MoMo. But it is not limited to small molecules. I think antibodies, siRNAs and CRISPR molecules (“off target effects”) also hit more than one targets in vivo.

      1. Mol Biologist says:

        Please do not confuse charisma with scrambled eggs. I would say that the drug only can work if it has a target. And all others multiple choices do not have anything in common with drug discovery.
        CRISPR is a bacterial system which design to eliminate all viruses and phages as enemies. The specificity is not highly required in this defense since it is targeting repetitive sequences or palindromes. Fooling is to pool everyting together.

  8. Peter Kenny says:

    This looks to be a very useful study in that it reports results for assays that have been designed to reveal bad behavior in assays (as opposed to frequent-hitter behavior). However, I would recommend that we stop using the term PAINS which is too poorly defined (even within the PAINS politburo) to be useful for anything but denouncing undesirable compounds as ‘enemies’. it’s important to remember that a compound can be thoroughly nasty without being especially promiscuous and a compound (e.g. adenine-mimic kinase inhibitor) can be promiscuous under typical assay conditions without behaving badly in an assay. It can be helpful to make a distinction between ‘active’ compounds interfering with assay readouts (while having no effect on target function) and compounds that affect target function by an undesirable mode of action. I’ve linked the ‘War and PAINS’ blog post as the URL for this comment and this takes a look at some of the developments in the PAINS saga.

  9. BernYeeIris says:

    How about ASS-PAINS Active Selective Scaffolds-Pan Assay Interference Compounds

    Ow, My Eyes!

  10. RBW says:

    The epigenetic community now prefers to call these proteins KATs (K for lysine) than HATs as there are plenty of nonhistone substrates.
    Going back to drug promiscuity, there’s a study by Jalencas and Mestres on approved drugs. About 50% had at least 5 targets with an affinity of 10 uM or better, and 10% had at least 20 targets.

    1. nowhere man says:

      And how many of those approved drugs achieve 10 uM or better concentrations in vivo for any length of time? (Let alone 10uM unbound concentration…)

      1. Peter Kenny says:

        On the subject of unbound concentration, the article (Smith &Li,The effect of plasma protein binding on in vivo efficacy: misconceptions in drug discovery, NRDD 2010, 9:929–939) that I’ve linked as the url this comment may be of interest.

  11. dg says:

    This paper and another one from Genentech (J. Med. Chem., 2016, 59 (23), pp 10549) describe excellent p300/CRB inhibitors. I don’t know what happens in companies (I work in academia); what does that mean for the internal development of these compounds?

    Have they given up on the target? Developed better molecules so they publish these as tool compounds?

    1. Barry says:

      Usually, such a project gets published when:
      -all the relevant patents have been filed
      – they’ve given up the target.

    2. toluene says:

      That paper describes antagonists of the bromodomain, NOT inhibitors of the HAT domain

Comments are closed.