Skip to main content


What We Can Do, Versus What We Could

I remember reading Barry Sharpless’ big “click” chemistry paper in 2001, where he proposed the term for reactions that take place rapidly, selectively, and without any outside reagents, and proposed such techniques for the rapid assembly of diverse molecules. In the years since, the term has drifted away a bit at times to mean “any reaction that forms a triazole from an acetylene and an azide”, but the concept is still a good one. And no one could have predicted where it would lead, almost twenty years later.

For example, it is (from what I can see) perhaps the central enabling technology of chemical biology. The ability to put a small unreactive tag on a biomolecule or small organic species and then label it at will, in vitro or in vivo, with whatever you like as a coupling partner is like having magic powers. It’s led to a vast amount of new knowledge about the protein landscape in living systems and their handling of small molecules. The versatility of the coupling reaction allows it to fit perfectly with other powerful techniques such as fluorescence microscopy, mass spectrometry, and photoaffinity labeling, and the amount of information generated through such methods is literally more than we can take in. Several of the people who’ve developed such experiments have been mentioned as candidates for a Nobel prize, along with Sharpless himself (for his second one!), and I would have no problem with that whatsoever. (There’s no space in this post to even go into the applications of click reactions in other fields; it’s a long and interesting list all its own and it keeps growing every year).

But as for chemical biology, this paper will serve to illustrate the sorts of things that you can do. It’s looking at bromodomain ligands, a field that everyone seemed to pile into at once a few years ago when small-molecule chemical matter was found to bind to the lysine-binding pockets on the proteins themselves. The lysines that they bind to are the ones that dangle off of histone proteins in the nucleus, and the bromodomain proteins themselves are involved heavily in the intricate dance of chromatin packing, DNA unwinding, and the incredibly complex process by which genes are selectively transcribed. (As an aside, this recent paper in Science has an extraordinary new look at the details of chromatin packing, thanks to a new electron microscopy technique).

Unraveling just which bromodomain-containing proteins are doing what, and when, and for what reasons and just exactly where is a tall order, but those are just the sorts of details that have to be tracked down to understand them. (Not to mention that such an understanding would greatly assist trying to turn any of the small-molecule inhibitors into drugs – it would greatly decrease the finger-crossing and breath-holding that accompanies any transition into human trials). The paper referenced in the paragraph above shows how to go about doing that.

The authors (from a multicenter collaboration in France and Australia) took several of the reported bromodomain compounds and added short chemical tags to them to participate in click reactions – either the classic azide/alkyne, or another later combination, tetrazine/trans-cyclooctene (an exceptionally fast reaction developed by the Fox lab at Delaware). Cell assays showed that these modified ligands seemed to behave just the same as the parent compounds, which is a key step any time you do one of these add-a-tag experiments. After using a variation of the crosslinking assays for chromatin structure, they could localize the compounds to particular DNA sequences by attaching a reporter group to them via the “click” handles.

This showed some interesting patterns that hadn’t been worked out before. As it turns out, the bromodomain protein under study (BRD4) behaves quite differently around different types of genes. It was already known that adding a BRD4 inhibitor immediately downregulates some genes, while other BRD4-associated genes don’t show this effect. This experiement showed that of the ligand accumulates at the downregulated ones, and the hypothesis is that the BRD4 protein interacts with them through its first bromodomain, leaving its second one free to pick up ligand. Meanwhile, BRD4 binding at genes where an inhibitor doesn’t have much effect seems to take place by a bromodomain-independent mechanism, probably a set of protein-protein interactions that not only don’t involve the lysines on histone proteins, but seem to make the lysine-binding sites less accessible to ligands.

Meanwhile, at enhancer sites in the genome, BRD4 inhibitors seem to generally downregulate things. Looking at these, some BRD4 protein-associated enhancer regions also showed noticeably greater accumulation of labeled drug than others. In this case, the binding does always seem to be dependent on the bromodomain pockets, and the differences may well be that in the high-ligand-amount sites the second bromodomain is left open (as above) to pick up compounds, while at the other sites both bromodomains might be occupied by lysines at the same time. (Alternatively, it could also be that binding to one of them is taking place in a way that makes the other one inaccessible).

Two different bromodomain inhibitors (IBET-151 and JQ-1) showed exactly the same pattern of accumulation in these experiments, which argues strongly that the two of them are mechanistically pretty much identical. They were already known to show very similar effects in cellular assays, but it wasn’t clear if they were getting to those endpoints via the exact same route, or if somewhat different compound mechanisms were just ending up with the same readout.

Another thing you can do with labeled drugs of this sort is to quantify their locations inside different cells, at different points in the cell cycle, and in different tissues. You can do some of these experiments on a less detailed level with radiolabeled compounds, but fluorescence microscopy gives you a real advantage here, not to mention being much easier to handle from the facilities and regulatory standpoints. It turns out that localization of the bromodomain inhibitors to the nucleus isn’t affected by the phase of the cell cycle or how proliferative the cells are in general. In mixing experiments, flow cytometry could be used to pull out the ones that had been exposed to higher levels of compound (another advantage of fluorescence). You could try doing that with a drug that’s already had a fluorescent tag on it from the start, but it’s much harder to do that and not affect its properties. A small click handle has a better chance of slipping by (as it does here), and lets you add the fluorescent group at a time of your choosing.

These studies were used to shed some light on an effect that had been noticed before, that bromodomain inhibitors, when used in models of acute myeloid leukemia (AML), seemed to do a much better job clearing out the cancerous cells in circulating blood and in the spleen than the ones in the bone marrow. Using the fluorescence accumulation in the nuclei as a marker showed that cells from the spleen and bone marrow can take up the compound equally, but the ones in the spleen just seem to be getting exposed to a lot more drug, which makes sense from a pharmacokinetic standpoint, since the spleen (technically speaking!) is perfused to hell and gone and access to the bone marrow is more limited. (The figures I’ve seen would indicate that, at least under normal resting conditions, blood flow through the spleen is at least ten times greater, mL/min per weight of tissue).

In general, this type of compound labeling, followed up by either fluorescence or mass spec detection (depending on the label you’d like to use) lets you track levels of drug between different cells types, different tissues, and down to intracellular compartments in vivo or even particular locations in the genome, if you’re targeting that machinery (as these compounds do). That’s a level of detail that just didn’t exist not so long ago, and a lot of it is made possible by these bio-orthogonal coupling reactions. Couple that with advances in handling genomic data and in protein identification (such as modern mass spec proteomics) and you can routinely run experiments that would once have gotten you laughed out of the room for even bringing them up. The process continues, too – we may well look back on these as the early days and feel pity for our younger selves for having to work under such primitive conditions. But for now, we can enjoy what we have and make the most of it!

16 comments on “What We Can Do, Versus What We Could”

  1. biotechtoreador says:

    I’ve always thought more of click chemistry as “reactions that have been around for decades that work well”, definitely useful, but I’m sure Huisgen in the 50s or 60s.

  2. executiveMBA says:

    I had to explain to a chemistry once that CuAAC is as truly regiospecific as you can get. Does anyone know if there is another clock reaction with the same regiospecificity? I don’t think there is, but I wonder where we could find it.

  3. road says:

    Not sure what this study learned that couldn’t have been gotten with a good BRD4 antibody…

    1. Derek Lowe says:

      Hmm – nuclear localization in live cells? And in general, behavior of these exact compounds, as opposed to tracking BRD4 by other means.

      1. sgcox says:

        Now, this is a part I do not fully understand – JQ1 and iBET work by displacing BRD4 from chromatin. The sites cross-linked by clickable analogues should therefore show chromatin sites where inhibitors do NOT work. That is the sequences where inhibitors bind to but do NOT displace BRD2/3/4. More informative about genes and enhancers affected would be simple BRD4 chipseg +/- JQ1. Pretty sure it has been done many times before.

    2. Blabla says:

      There’s probably more proteins out there that don’t have a good antibody than those that do, so that’s a big if…..

  4. Mfernflower says:

    Everything old is new again!

  5. Hopeful says:

    If only somebody could figure out how to use these click chemistry tags for in-vivo therapeutics and improve biodistribution

  6. jb says:

    Actually, you don’t even need click chemistry for azido or alkyne groups to be useful in biomolecules, alkynes are Raman active:

  7. PotStirrer says:

    I thought another important aspect of a “Click” reaction as per Sharpless’s definition was that it was an addition reaction…no byproducts.

  8. Imaging guy says:

    As a person who is interested in pharmacohistology (i.e. in situ detection of injected small molecule drugs and therapeutic antibodies in tissue sections), I have certain misgivings about this article. Fluorescent microscopy images of the clickable drugs [(JQ1-TCO, IBET-762–TCO Fig 3B) and (JQ1–PA Fig S8.D] in in vitro cell cultures are similar to those of BRD4 immunohistochemistry, which is the intended target protein. They are all in the nucleus. So, this part is fine. However, the image of femur bone marrow tissue section taken after in vivo injection of clickable drug JQ1-TCO [Fig 4D and Fig S8.F (control)] does not seem to show specific staining pattern (images for spleen tissue sections are not shown). No images of BRD4 immunohistochemistry is shown though it could have been easily done on adjacent tissue sections. BRD4 immunohistochemistry gives very specific staining pattern in tissue sections as you can see from these Human Protein Atlas (HPA) images (1). The point is clickable drug JQ1-TCO should also produce similar specific images in tissue sections. [Although enzymatic chromogenic stain instead of fluorescent ones are used in HPA images, similar chromogenic images of clickable drug JQ1-TCO could have been easily obtained by using Biotin-Tetrazine (which they used for pull-down) and streptavidin-HRP enzyme which is widely available.] My conclusion here is clickable drug JQ1-TCO binds non-specifically in vivo. I don’t know whether this is also true for the parent drug JQ1. I have seen this kind of non-specific binding in tissue sections after in vivo injection of other clickable drugs. Here you might wonder what kind of images you obtain with labelled therapeutic antibodies in vivo since they are supposed to be more specific than small molecule drugs. They do seem to produce specific staining pattern as you can see in this article (2).
    2) Improved decision making for prioritizing tumor targeting antibodies in human xenografts: utility of fluorescence imaging to verify tumor target expression, antibody binding and optimization of dosage and application schedule (PMID: 27661454)

    1. Doug Johnson says:

      I have dabbled in this area and I agree nonspecific binding can be a big problem. It is really important to do the proper controls (like competition experiments and images with just the fluorophore) some of which were missing in this paper.
      Nonetheless, I think this is an emerging area and can benefit from new fluorogenic fluorophores.

  9. Andy II says:

    In the early days back in 2002, Sharpless and Finn at TSRI reported a very intersting fentamolar inhibitor of acetylcholinesterase thru an in situ click chemistry in AngewChem,
    2002, 41, 1053-7. Since then, there are many publications from mostly academic labs using a similar strategy to come up with compounds with residues that bind multiple binding sites. I don’t know if these chemistry-biology would have produced druggable molecules that are in someone’s pipeline…

  10. Jb says:

    There are also ‘click activated’ fluorophores that only fluoresce once clicked. In theory it is possible to have 0 background using such an activatable fluorophore, or with using Raman microscopy to detect alkyne tags. When are antibodies ever 100% selective with no background?

    1. Barry says:

      alas, lots of WT proteins fluoresce; rats fluoresce. There’s always background. But if you use a lanthanide fluorophore with time-gating, you can get the background way way down

      1. JB says:

        That’s why the Raman microscopy approach in the article I linked to above is so neat. Natural molecules and tissues aren’t usually Raman active. Installing alkynes and other Raman active groups achieves virtually 0 background. Tissue penetration is also deeper w/ Raman than typical fluorescence detection .

Comments are closed.