Skip to Content

Chemical Biology

Only Bind

Here’s a new piece by Stuart Schreiber that lays out a shift in thinking that many people in the chemical biology field have been experiencing over the last few years. Medicinal chemists are used to making functional drug candidates – and by “functional” I mean compounds that make protein targets do something. For an enzyme, that usually means an inhibitor, because it’s always easiest to gum up the works, and the same goes for antagonists of receptors. But you can, of course, have full agonists, partial agonists, inverse agonists and all everything in between, and for enzymes you can have allosteric inhibitors and (rarely) activators as well. But what all of these have in common is functionality.

A great many assays are set up to detect such functional changes, naturally, looking for changes in some output in the protein or downstream in a cell. But you can also have compounds that bind to some protein without really doing much to its activity. Those will be silent in such functional assays, but you can pick them up with biophysical techniques like NMR, SPR, etc. DNA-encoded library screens also pull out binders, rather than screening on function. But in the past, having a binder that didn’t do anything didn’t necessarily get you very far – a difference that makes no difference, etc. What people would do with the results of such assays, likely as not, was to funnel them right into a lower-throughput functional readout and toss the compounds that did nothing. Which was sometimes all of them.

That’s changing, though. The most obvious technique working this way is targeted protein degradation. If you have something that binds to your target of interest, even if it doesn’t affect it by itself at all, you could imagine using that as a handle to engage the ubiquitination/proteasome machinery to drag the protein itself off to the compost heap. But it’s become increasingly clear that small molecules can (at times) cause two proteins to stick together. Consider the poster child for this mechanism, rapamycin: it binds tightly to FK506-binding proteins, and the resulting surface turns around and binds to mTOR proteins/complexes. But that’s a dramatic example, and it doesn’t stop there. Just binding a compound to a protein somewhere may not seem to affect that protein’s direct functional readouts, but (on closer inspection) it could alter its conformation or the surfaces it presents in ways that affect binding to its various partners.

These may well not be obvious – perhaps they’re not happening anywhere that your functional assay is looking, or only happen under certain conditions. Such binding events could also affect stability or localization of proteins as well, or their (many) posttranslational modifications. It’s not surprising that it’s taken years to become aware of such things and to start to take advantage of them, because we really haven’t had the tools to look at such broad ranges of behavior for very long. The number of possible interactions that a given protein can make is pretty intimidating – for some of them, it’s downright terrifying – and these can (and do) vary by cell type, environmental conditions, phase of the cell cycle, and so very much on.

Schreiber’s article goes into these and more, and suggests ways that the drug discovery/chemical biology/cell biology community might go about looking for these things systematically. This is a large landscape, and we know zilch, to a good approximation, about a lot of it. But it’s absolutely worth exploring, because we may be able to find ways to attack targets and pathways that until now were considered ridiculously out of reach for small molecules. So all you small-molecule folks: here’s the frontier, and here’s how you can find yourself back in the highlands, surveying unknown territory. Come and get it!

20 comments on “Only Bind”

  1. Wavefunction says:

    Measuring binding is at the heart of DNA-encoded library technology because of its use of affinity screening. But empirically, many of the binders do tend to have functional relevance.

  2. luysii says:

    On a much larger playing field, the immunologic synapse with its multiple interactions between the Major Histocompatibility Complex bound to an immunogenic peptide (pMHC) and the T Cell Receptor (TCR), the Kd does NOT correlate with the biologic effect your interested in (e.g. the induction of an immune response). Something unknown in solution chemistry (the catch bond which strengthens when pulled on — like the seat belt in your car) is involved.

    For details read — [ Cell vol. 174 pp. 672 – 687 ’18 ].

    Or read – https://luysii.wordpress.com/2018/08/05/when-the-dissociation-constant-doesnt-tell-you-what-you-want-to-know/ — if you don’t have a subscription.

    1. Imaging guy says:

      Peptide-MHC binding to T cell receptor (TCR) is quite weak and does not usually survive washing step. In order to identify T cells which bind to different peptide-MHCs with flow cytometry or in situ staining, tetrameric peptide-MHCs are made with streptavidin. There is even a NIH tetrameric core facility.
      1) “Interrogating the repertoire: broadening the scope of peptide–MHC multimer analysis”
      Nat Rev Immunol. 2011 Jul 15;11(8):551-8. doi: 10.1038/nri3020.
      2) “Combinatorial tetramer staining and mass cytometry analysis facilitate T-cell epitope mapping and characterization”
      Nat Biotechnol. 2013 Jul;31(7):623-9. doi: 10.1038/nbt.2593

      1. luysii says:

        Imaging Guy — Quite true — the article notes that the kDs are in the microMolar not nanoMolar range. The second article referenced in the post — [ Proc. Natl. Acad. Sci. vol. 115 pp. E7369 – E7378 ’18 ] cleverly added an extra few amino acids (which they call molecular velcro), to boost the affinity x 10 (actually this decreases Kd tenfold).

        One rationale for the weak binding is that it facilitates scanning by the TCR of the pMHC repertoire allowing the TCR to choose the best. So they added the velcro, expecting the repertoire to be less diverse (since the binding was tighter). It was just the same. Again the Kd didn’t seem to matter.

        1. Immune says:

          This relative indifference to binding affinity even extends to T cell eradication of cancer, at least in some cases [Transnuclear TRP1-specific CD8 T cells with high or low affinity TCRs show equivalent antitumor activity. Cancer Immunology Research 2013]

    2. Bla says:

      Argh, every time someone misuses the term synapse like that, another part of my neuroscience heart dies.

      1. luysii says:

        Bla — Hey I’m a retired neurologist, and was a board examiner at that, but if you want to communicate you have to use terms that everyone uses.

  3. Not going gently says:

    Definitely support your call to arms on this Derek. It’s tempting to bet that directly measured functional activity is only part of the biological impact of most small molecule drugs. We won’t find what we don’t look for – it’s a failing of our discipline that we limit our perspectives in this way.

  4. McDELT says:

    Bilkions and billions screened 🍔

  5. David Edwards says:

    As someone with an involvement in invertebrate zoology in the past, the mention of rapamycin and FK506 rang a bell. Turns out rapamycin and FK506 were used to investigate the underlying molecular biology of diphenic caste development in honey bees – this paper covers the details. So the paper Derek’s presented here may not only have use in med-chem, but in other biological research as well. Find a small molecule that interacts interestingly with a system under study, and whose interactions therewith make the light bulb go on over the head, so to speak, as to what’s happening in that system, and we move forward, even if the research in question wasn’t originally med-chem directed, because it could give the med-chem people some interesting ideas as to how to proceed with a problem they’re working on.

    Now if someone can come up with a small molecule that helps us understand what’s really happening in Alzheimer’s, even if said small molecule doesn’t have therapeutic value in itself, that will be a welcome step forward. Though I suspect alighting upon a rapamycin/FK506 analogue for investigating the behaviour of Alzheimer’s will itself be a hard problem to solve. But in the light of this paper, and that prior work on bees, I suspect this might be an approach worth looking at – find a small molecule that illuminates our understanding before trying to solve the harder problem of therapeutics.

  6. drsnowboard says:

    A Schreiber Nobel Manifesto?

  7. john adams says:

    Of course Schreiber is NOT the first scientist to recognize that binders can be identified by screening, including very high throughput methods (e.g., ALIS). The issue has always been “what do I do with these compounds once identified and shown to NOT work in my (favorite), and usually in vitro/biochemical, functional assay(s)”? While I have yet to read the article (and so this may have been suggested), I’d like to propose that (reasonably potent/cell permeable) binders be examined in phenotypic, whole cell based assays for activities of interest (including those relevant to the initial molecular target, if known). In the (rare?) cases where an interesting activity is found that is clearly not toxic (e.g. as measured by cellular ATP levels), then at that stage it would likely be worthwhile to look at its effects on protein-protein interactions/post-translational modifications/etc. of the molecular target of the binder.

  8. Free Radical says:

    Slightly tangential on this FK506 thing, but can anyone tell me what the “FK” stands for? Pretty sure the “F” is Fujisawa, but why “K”? Kagoshima? Pretty sure “FR” would be Fujisawa Research. Derek’s very old post on this is still useful:
    http://blogs.sciencemag.org/pipeline/archives/2006/10/23/experimental_compound_codes

    Also, this is related to Bradner’s recent “dTag” (degradation Tag) thing using FK binders to fusion proteins made using Crispr. Seems like a good approach if you don’t have a ligand for the protein you want to degrade. Hard to make the molecules, until someone starts selling them. https://www.ncbi.nlm.nih.gov/pubmed/29581585

    1. d says:

      I’m not sure on the name, but you can start with SLF (commercially available) to generate bifunctional ligands that bind to FKBP. As long as your ligand of interest has a handle amenable to linking to SLF, the chemistry is actually not all that bad (made many of these in my thesis work).

    2. CA prof says:

      Fujisawa compound codes: The research-level compounds were marked “FR”, usually with a longer numerical sequence in front. So, FK506 would have started out at FR900506. Once it moves to the “klinical” stage, then it goes to “FK” status.

  9. AnonymousAgain says:

    In my experience, if you didn’t find the right number or kind of binders in the first pass, you just go back to the data and adjust the threshold until you get what you need.

    You don’t want a library with 90% hits – too indiscriminate. You don’t want a library with zero or one hit – suggestive of too much garbage. Just check to see what management put on their slide presentations and dial in the cutoff until you hit their mark.

  10. Larry says:

    Have not read Schreiber’s commentary so perhaps I’m observing an opportunity already noted, but an obvious application for the molecular glue (or velcro) approach is for small molecules that will specifically & selectively increase the affinity between substrate proteins which are targets for post-translational modification (PTM) and those enzymes (e.g., protein kinases or phosphatases) responsible for catalyzing the PTM. In many cases, those PTM enzymes have low intrinsic affinity for their target substrates (operating in the kcat/KM range), but such a small molecule could block the action of a PTM enzyme as effectively as a more conventional inhibitor.

  11. Grad student ( going to be cool oneday ) says:

    When’s the last time we have seen a b cravatt post. Looks like dupity Derick found a new love: Stewart scheiber, lol.

  12. Matt S. says:

    My boss and I have a running friendly wager on which is more common: A) Binding compounds with no functional readout or B) Functional compounds with no reported binding.

    1. Scott says:

      I’m afraid to ask who is leading in that pool…

Leave a Reply

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

Time limit is exhausted. Please reload CAPTCHA.