Time for a word about screening for new coronavirus drugs. Things have gone on long enough for quite a few groups to produce supplies of the various viral proteins and set up small-molecule screens against them. That’s no bad thing in itself, although it is a slow thing, a very slow thing by the standards of what people are looking for in the epidemic. Screening a new protein against a collection of small molecules is one of the first steps along a pretty lengthy road – you’re looking at several years if everything works perfectly.
That’s a big reason why that the small-molecule antiviral drugs that we have (which is not a large set compared to many other therapeutic areas) are very much skewed towards long-running viral diseases like HIV and hepatitis C. Those have had a time scale that matches with drug discovery, especially when you consider that really effective small-molecule antiviral therapies tend to be cocktails of several drugs hitting several targets simultaneously. Coming up with a whole new suite of drugs is no small order.
Contrast those diseases with things like SARS and MERS, which (fortunately) vanished by the time anything in the small-molecule line could be developed specifically against them. Even Ebola, which has clearly been a longer-lasting problem, has still messed around with clinical development efforts. Remdesivir, famous now because of its possible utility against the coronavirus, was originally identified during the 2013-2016 Ebola outbreak, but then faced a shortage of human patients for efficacy trials. Unfortunately, the re-emergence in 2018 provided more opportunities, which actually led to the discovery that the drug was not as effective as the monoclonal antibodies that had also been developed. That’s why I’ve been highlighting the drug-repurposing efforts when it comes to the small molecule therapies, because those are ready to go into humans right now. Fresh screens are not uninteresting, but their impact is just not as great and cannot be as great, especially for the current outbreak.
Most such efforts have focused (very reasonably) on the “main protease”, often known as Mpro. Many viruses have such a thing, in a bootstrapping sort of mechanism – a virus will force cells to crank out one or two viral protease enzymes along with some long protein chains, whereupon the proteases start cleaving the long stuff down into other functional viral proteins. It’s a neat process if you could just watch it from afar, in the cells of some other species. The SARS coronaviral main protease, for example, cuts at 11 different sites in a complex sequence of events that’s like watching a robot assemble itself by unfolding from a small shipping crate. In the case of a virus, though, that robot proceeds to take over the rest of the equipment in the house to make more shipping crates, leaving the place totally destroyed and blowing the new supply of crates all over the neighborhood.
Here’s a paper from a large team in China who came together with commendable speed to obtain the X-ray crystal structure of the new coronavirus Mpro enzyme. Note that although this was published in Nature on April 1, it was received on February 9 – as I say, these folks got moving quickly. Here’s another structure from a joint German-Chinese collaboration, published on March 20. Earlier in March another group at the Diamond synchrotron facility in the UK also obtained the protease structure in collaboration with the Chinese research teams and ran a fragment screen against it, which they have expanded to a crowdsourced effort to find new inhibitors.
I should be more into this work than I am, when I consider how much I like structural biology, fragment screening, etc. But as you can tell from the above paragraphs, I find myself so loaded with immediate concerns that the beginnings of longer-term work that might bear fruit in a few years just keep moving down the list for me. My longer-term hopes for therapies against this virus lie in vaccines (especially) and perhaps monoclonal antibodies, and both of those will read out, multiple times, before the first molecules that eventually come from these screening efforts will make it into human patients. That actually makes these small-molecule efforts more important in a way, though, because if they eventually become useful, it will be because those other attempts at treatment have failed, and then they will be very important indeed. I very much hope that we don’t get to that point. It’s a psychological problem for me: initially I react to these screens with doubts about their utility, and then if I picture the situation where they are indeed useful I react with horror.
Now to some practical considerations for the folks doing this work and those reading about it. These viral enzymes tend to be cysteine proteases with a busy, nucleophilic SH group in their active sites. And that leads to some problems when you screen against them. As anyone who’s done small-molecule screening knows from experience, There Are Some Compounds (and some compound classes) that just tend to hit more often than others. Call them PAINS, call them any nasty name you want – you’ll have plenty of opportunities to see them, because they will be back around again. With cysteine proteases, you will be seeing electrophiles that can react with the SH group, and plenty of them. Such things will indeed shut down the enzyme, but they are likely to shut down a lot of other stuff, too.
You can see that in the structures that are shown in Figure 3 of the Nature paper. That’s quite a crew of known drug compounds: you have ebselen (which I wrote about here) and which hits so many things that we’re still trying to figure them all out, disulfiram (mentioned in this post as “a shotgun” of a compound), tideglusib, a covalent inhibitor of glycogen synthase kinase 3 that has failed a clinical trial in Alzheimer’s (no great distinction, everything else has too), carmofur, an investigational oncology drug that also binds covalently and was pulled from clinical trials due to adverse events, shikonin (that link is to its enantiomer, alkannin, but they’re both hideous deep-red napthoquinones that will react covalently and do all sorts of redox chemistry to boot), and PX-12, a thioredoxin inhibitor that has failed in the clinic and is another reactive disulfide. These are just the sort of things I would expect to see in a cysteine protease screen, and to be honest, they are pretty much the wretched refuse of the teeming shore.
That description goes for a lot of the compounds that have been suggested so far through physical or virtual screening. The comments here have been filling up with people from outside the field talking about quercetin, for example, and the reason that I know that they’re from outside the biomedical field is that they’re talking about quercetin. That’s a classic little polyphenol flavonoid found in all sorts of plants. It has wonky pharmacokinetics (but little or no toxicity) and it has been through all sorts of small trials, hits in all sorts of assays all the time, and has rarely been found to be particularly useful for anything. Honestly, if you come up with quercetin, ebselen, disulfiram and the like as your screening hits it just means that you have to keep looking. This is no secret – the various groups doing these screens know this as well as anyone, and they are indeed looking further. But under the current conditions, when these things show up in the literature, people who have not been around these particular blocks so many times can have a tendency to jump on them as The Latest Hope – it’s understandable, for sure, but it’s not doing anyone any good.