I wanted to mention this paper, which is one of the more comprehensive ones on the idea of repurposing existing drugs against the coronavirus. It’s a large multicenter team that clearly did a lot of very fast coordination to produce these results. What they’ve done is looked at the complete suite of proteins produced by the new coronavirus (some 27 to 29 of them, we think – viruses have relatively few moving parts). They were able to express almost all of them with a strep-tag in human cells (HEK293T), whereupon they used affinity-purification mass spectrometry (APMS) to identify human proteins that associate with them. With that technique, you lyse the cells and run that over beads containing a particular engineered form of streptavidin that the strep-tag binds do.. The strep-tagged proteins stick to the beads, along with whatever proteins are sticking to them, and then you clear these off by eluting with biotin (which binds to streptavidin like nothing on earth) and analyze them through mass spec proteomic techniques. It’s a classic chemical biology experiment, and works pretty well (with some known artifacts, naturally).
Looking at the interaction profile of 27 of the viral proteins, with suitable control experiments for nonspecific binding, etc., gives you 332 total interactions with the human proteome (well, the human proteome as expressed in HEK293T cells, anyway). To that point, the team went on to check the expression of these human proteins across 29 different tissue. Most interestingly, lung tissue (which is of course a lot more relevant to the infection!) had the highest expression of the overall suite of hits, which certainly argues for their functional importance in the viral mechanism. To go along with this, there’s recently been a report of protein expression changes during viral infection itself, and indeed, the interacting set also stands out in this group. In fact, a number of them also show up as hits in similar protein expression screens that had previously been conducted in West Nile and tuberculosis infection, both of which of course are also associated with lung pathways. So these pathogens may well have converged on some likely targets over time.
Running the human proteins through a Gene Ontology enrichment analysis to try to sort out functions showed some likely targets and pathways for several of the viral proteins – for example, the viral Nsp7 seems to be involved in nuclear transport processes. As you would expect, there are a lot of signs of the virus attacking the pathways of innate immunity and inflammation. You can even go down to the protein structural domain level and find enrichments – as an example, the viral Nsp1 protein’s human interactors are enriched in DNA polymerase domains, and the viral E protein’s interactors are enriched in BET and acetylated-histone binding domains. In addition to these sorts of transcriptional targets, some of the processes targeted by the viral proteins include vesicle trafficking, lipoprotein pathways and lipid metabolism, and mitochondrial and cytoskeletal proteins. There are several potential targets in these that could also apply to other pathogens, which is good to know in the long term.
For the shorter term, though, the team tried to identify known compounds that interact with the human target proteins, in the hopes that some of these might go on to interfere with the viral protein-protein interactions that had been uncovered. There are about 25 approved drugs on the list, which are very likely of the most interest, since these can already be dispensed to patients. It’s an interesting list. Chloroquine shows up, but not by any of the mechanisms that anyone has proposed for it – it makes this list via its capacity as a Sigma1 binder, of all things. If it turns out that sigma receptor ligands are finally good for something I will be quite startled, since I’ve been hearing comments about ignoring sigma activity because everything hits it and no one knows what it does since, oh, about 1990 or so. Azithromycin didn’t make the list per se, but it is known to have off-target ribosomal effects, and compounds with more direct ribosome targeting were identified.
And the ACE inhibitors are on there, too, which brings up a key point. If you’re just looking at the level of “This compound interacts with this human protein, which apparently interacts with this viral protein”, which is all we can do for now, that doesn’t give you enough mechanistic detail to say if said compound is going to do anything beneficial, to turn out to have no effect, or indeed to possibly make things even worse. This is a very nice paper, but the authors themselves are careful to note that an interactome map like this can’t provide any mechanistic understanding by itself – we have to bring that, via brainpower and further experimentation. Some of these proteins and pathways are being hijacked by the virus or interfered with by it, while others may be involved in fighting it off, so you need to be sure what you’re messing with.
There are also a lot of compounds that are in clinical trials that show up on the list, which are one notch down compared to approved drugs, and a number of other compounds that are known in the literature as ligands, tools, etc. but have not been into humans yet. Those are way down in the bottom rank as far as I’m concerned. The big hurdle is getting drugs into patients, and having been through the preclinical studies needed for human trials (toxicity, dose formulation, stability, etc.) and especially having been into Phase I or II trials in actual human subjects counts for a great deal here. All the better if the compound proved safe enough (and efficacious enough in its intended role) for a regulatory agency to approve it. Those are actionable; pre-clinical compounds are far, far less so.
This paper points out several things that need to be followed up on, and I’ll be going into more detail on those in future posts and as more information comes out. For now, congratulations to this big team for moving so quickly and getting this out to the research community!