5-fluorouracil (5-FU) has been around a long time now (over fifty years), and it’s a standard oncology drug (particularly in colorectal treatment regimes). But try going around and asking people how it works. If you’re talking to a clinician and want to seem up on the lingo, just say “What’s 5-FU’s MOA?” (mechanism of action).
If your interviewees are honest, many of them will say “Actually, I don’t know”. The others will likely tell you that it’s via inhibition of thymidylate synthase, interruption of which leads to DNA damage. There’s no doubt that 5-FU is indeed a thymidylate synthease inhibitor, and there’s no doubt that that can mess up DNA synthesis, but is that all that’s going on? After all, this is a pretty small molecule that resembles a very important cellular building block. It’s long been suspected that there are other effects.
This new paper actually sheds some light on this topic, and people should prepare for some surprises. There have been many studies before on 5-FU’s cellular effects, of course, but many of these have been at the protein expression level. That’s valuable, but it can miss a lot. Not everything that changes gene transcription is part of a drug’s actual mechanism, because a lot of drug mechanisms don’t go through changes in protein expression. Rather, they set off changes in protein-protein interactions, directly or indirectly. Expression profiles can be sideshows, things that happen because of the main mechanism. Those readouts can be well downstream, mechanistically, and can take much longer to come on than the actual MOA. You can get valuable hints about what might be going on, for sure, but you can also be misled. Taken to an extreme, one risks concluding that piles of wet smoldering ashes are the main thing that causes damage when a house catches on fire.
The authors (a multinational team from Sweden, Singapore, Germany, and England) used the CETSA technique to try to get a more direct mechanistic readout. I’ve written about this a few times here, but the basic idea is that when a protein interacts with a binding partner, it gets slightly more stable. And you can actually see this by heating it up gently and seeing when it starts to unfold – in a bound state, it will do so at a higher temperature than it does in the unbound state. This can be done in vitro by adding a fluorescent dye that reads out on the unfolded state (the DSF assay), and there are other techniques for this “thermal shift” assay as well. What became apparent a few years ago was that you could do this in living cells as well, which opens up a way to get a lot of interesting information that’s otherwise not easy to obtain. This paper uses the latest version of this cellular thermal shift assay (CETSA), which uses protein digestion and mass spectrometry as a readout, and it’s a good advertisement for its power.
The study wisely decided to start by using the active metabolites of 5-FU (5-fluorouridine, FUR, and 5-fluorodeoxyuridine, FUDR) to get the most immediate effects, exposing the cells for two hours before doing the CETSA heating and mass spec data collection. They found several proteins that had already been noted as affected by the drug, which is a good reality check, but they found many more that had never surfaced at all. Interestingly, both FUR and FUDR showed dozens of proteins that read out differently in the thermal shift assay, but there were only seven common hits between the two of them!
As you can see from the names, it’s quite likely that FUR would be going down RNA pathways, among other things, while FUDR would be acting through DNA ones. Indeed, the proteins on the latter one’s list were biased towards nucleoside processing pathways and DNA repair, which suggests that the thymidylate synthase pathway is probably being affected, and that the fluorodeoxyuridine itself may well be getting incorporated into newly synthesized DNA, which quickly attracts repair enzymes when things then gum up.
Meanwhile, the FUR list were largely RNA-handling proteins. These include a whole family of pseudouridine synthase enzymes, two of which had been shown in earlier studies to be affected in nematodes and other species on 5-FU treatment, so that’s a good thing to have picked up again. In the same category is TRMT2A, an enzyme which is believed from other work to form a covalent adduct with transfer RNAs bearing fluorouridine on them, and which has been implicated in 5-FU toxicity. There were also enzymes involved in dihydrouridine synthesis, and there’s a common feature in all of these categories: all of them work on the 5-position of the heterocyclic ring. That suggests that FUR might well be forming covalent adducts in the active sites of many of these, thus leading to a strong thermal shift readout.
It needs to be mentioned that those two-hour cell experiments mentioned earlier involved virtually no changes in protein expression levels (it’s too short a time to see very much of that). It’s all protein-protein interactions that drive these effects – you’d never pick these up by looking at expression levels, and indeed, no one had.
When the authors tried treating an actual colon cancer cell line (HCT15) with plain 5-FU for 12 hours (to simulate real-world conditions), it turns out that the response to the drug seems to be mostly in the FUR-affected proteins. That is, the biggest response to 5-FU treatment is in RNA pathways, not the DNA ones that would be suggested by the putative mechanism of inhibiting thymidylate synthase. In fact, a whole group of tRNA ligases showed up only under this sort of 5-FU treatment (as opposed to using either of its active metabolites). A further experiment used HCT15 cells that had been exposed to enough 5-FU over time to become resistant to it. This experiment showed that the thymidylate synthase pathway seemed to be largely unaffected in the CETSA results, while various RNA-handling protein hits were now changed. Other proteins that showed strong changes in the resistant cells were enzymes like UMP synthase, which are themselves involved in activating 5-FU, so seeing those moving out of the picture is just what you’d expect in a resistance environment.
Taken together, all this suggests that we have been mistaken about the real mechanism of 5-FU. While it does indeed inhibit thymidylate synthase, the more important cellular effects are on the RNA end of things. The authors propose a short list of proteins that could be used in analyzing biopsy samples to predict 5-FU sensitivity (or conversely, to indicate that the drug might be unlikely to work).
This is why I like chemical biology. Opening the hood on mechanisms like this is not easy work, but it’s essential if we’re ever going to put our therapies on a more rational basis. As it stands, we give 5-FU to patients with a mistaken idea of what it’s doing, and when it stops working we really don’t know why. Work like this (and the many other studies that have tried to unravel 5-FU mechanisms in the past) is the only way to change that.