One of the marquee chemical biology techniques is in vivo labeling of proteins and protein binding sites. There are vast numbers of interactions going on in a living cell, many of them rather transient (but still very important), and figuring out just what is meshing with what, and when, and why, is obviously going to be crucial to understanding the biology of both healthy states and disease. You really have to do that in situ, difficult as that is, because it’s really hard to recapitulate the cellular environment anywhere else. Cell lysate (break ’em open and mix up the contents) is the next best thing, but if you end up just testing individual proteins in vitro you’re surely going to be misled a lot. An artificial system risks having serious differences in the state of the proteins themselves (phosphorylation and other post-translational modifications), in the relative concentration of the partners, the presence of other proteins that could be involved in the complexes, in the properties of the medium (ionic strength, viscosity, crowding) and many more.
But doing labeling in living cells is no stroll through the tulip patch. If you’re looking at where small molecules are going, then photoaffinity techniques are the way to go (that means putting a group on your molecule that will turn into a ravenously reactive intermediate on exposure to the right wavelengths of light, whereupon it will form a covalent bond with whatever it’s sitting next to). There are quite a few of these, but the diazirine (a three-membered ring with an N-N double bond in it) is probably the first choice, because it’s so small. You can really throw things off if your molecule has to be modified with some bulky thing (losing binding activity against the desired target(s), picking up binding to other unwanted things, throwing off localization and distribution, and so on).
And if you’d like to know which proteins are in the vicinity of which other proteins, there are a number of options, too. One popular one is BioID, which uses an enzyme that will slap a biotin group onto most any protein it encounters. You take your protein of interest (POI) and fuse it to this bioten ligase (first making sure that it will still behave roughly as you expect it to!) and then turn it loose in the cell. The proteins that associate with your POI will get biotinylated as well, giving you a classic handle to isolate them from other cellular components and to compare your catch with control experiments. BioID is good stuff, but it has it limitations. A big one is that it works over a time scale of hours, so you’re going to be getting a big-picture look at protein-protein interactions rather than a fine-grained one. That’s led to faster versions of the technique, ones that can be better localized and so on, and new stuff in that area appears all the time.
A couple of years ago, the MacMillan group at Princeton introduced another new technique for such studies. It’s sort of a combination of the concepts behind BioID and classic photoaffinity labeling. It features an iridium catalyst that can be hooked to a number of different species (antibodies, small molecules, proteins, oligonucleotides, carbohydrates), and that catalyst is matched to the reactivity of a trifluoromethyldiazirine species. When you shine blue light into this system, the Ir catalyst activates that diazirine (through a process called Dexter electron transfer), causing it (as with traditional photoaffinity species) to turn into a highly reactive carbene that will label whatever it’s next to. You can see the first improved thing about this system: you can (if you wish) localize both parts of the system, the catalyst and the diazirene, and zero in on cellular processes more closely than many other techniques allow. Another advantage is that the Dexter energy transfer only happens at very short ranges, making things even more precise.
Since this work appeared, there have been several interesting applications of it and of related techniques. For example, here’s a small-molecule one where several known drugs and probes are profiled in cells with the iridium catalyst technique (known as µMap), revealing both their expected targets and a range of other binding partners that were largely unknown. And here’s a different technique entirely, conceptually similar but using flavin cofactors to produce reactive phenoxy radicals (PhoMap), used in this case to map the interactions between entire cells (immune cells and antigen-presenting ones, for example). The whole photocatalyic labeling idea clearly seems to be taking off, and you’d have to imagine that a lot of other experiments are in the works right now that we haven’t heard about yet.
And now there’s an extension of MacMillan’s catalyst idea, “µMap-Red”. This one uses a tin(IV) complex that absorbs red light and can activate a phenyl azide probe molecule, in this carrying biotin on it for later use after labeling. They demonstrate this in vitro with a test protein (carbonic anhydrase, everyone’s favorite platform), and go on to label EGFR protein on cell surfaces, with the tin catalyst being direct by conjugating it to appropriate antibodies. They then take full advantage of the tissue-penetrating properties of red light by using TER119 antibodies to direct the catalyst to the surface of red blood cells (it’s well-known to pick those out in mice and is used as a reagent in several assays for that purpose). Adding the phenyl azide bioten probe to a vial of whole mouse blood and incubating this with the antibody for an hour, followed by ten minutes of red light, gave robust labeling of erythrocyte surface proteins with biotin. Trying the same sort of labeling experiment with the original blue-light µMap experiment, though, gave no labeling at all, since blue light can’t penetrate!
The obvious next step is to try to get this to work in a whole animal, and you can bet they’re beating away on that right now. That’ll be more of a challenge – the paper itself lists several problems to be overcome (pharmacokinetics, local heating, phototoxicity, delivery of the needed NADPH reductant in this catalyst system. Those in fact make me think that they’ve already tried it and found that it would be better to publish what they have! But this is a real advance and a sign of progress to come.
A huge amount of data will be generated using these various labeling schemes and the ones that will follow on this work. It’ll be great: these could be bright windows opened on what has until now been just darkness, inference, and speculation. Another step in finally understanding what’s really going on!
