The last post talked about antibodies to the spike protein of the coronavirus, and one of the main things that everyone has to keep an eye on are the mutations in that area. That has implications for monoclonal antibody therapy, for vaccine production, and for the behavior of the coronavirus itself. Antibodies against the Spike protein could range from neutralizing ones that will stop the virus in its tracks all the way to others that would cause antibody-dependent enhancement and make the viral infection even worse (see below), and we don’t know how the mutational landscape might alter the activity of any given monoclonal candidate. A new preprint on spike muations (from researchers at Los Alamos, Duke, and Sheffield) has gotten a great deal of attention in the last couple of days, and I think that a detailed look at it would be useful to help explain these issues.
Most of this manuscript focuses on a particular mutant called D614G. For those outside the field, that’s a standard notation that translates as “at amino acid residue 614, take the existing amino acid (aspartic acid, single-letter code D) and change it to a glycine (single-letter code G)” That change happens through the triplet code in the RNA sequence that codes for the amino acids – a single random-mutation A-to-G switch in the right spot is enough to flip the resulting amino acid from an aspartate to a glycine (GAU and GAC both code for aspartic acid, and GGU and GGC both code for glycine). And those are pretty different beasts: aspartic acid has a polar, charged side chain, while glycine is the no-side-chain-at-all generic amino acid. This is not expected to be a silent mutation, in other words.
It’s important to realize that this mutant did not just suddenly come into view in the last few days. In fact, readers of this blog have encountered it before, in this post on coronavirus mutations back on April 21. Which of course seems like about six months ago, but you all know how that goes, right? (That one might be good background for this post, if you haven’t read it). I reproduced a graphic from the Nextstrain.org folks that showed mutations in the virus’ sequence in terms of entropy – that is, how different the amino acid residues were from the canonical sequence. Here it is again:
As that earlier post explains, moving from left to right you have the whopper ORF1, which codes for the equally whopping polypeptide that is the self-unzipping archive of a lot of the key viral proteins (it really is alarming how these things assemble themselves; there’s no easy analogous example out here in the macroscopic world that gets the weirdness of it across). Then you have the green region marked “S”, and that’s our friend the Spike protein. Note right smack in the middle of it is that big tall line: that is the G to A switch that gives you the D614G mutation itself, and it’s so tall because that entropy measurement is the degree to which the amino acid has changed. So people who are into coronavirus mutations (I mean, who isn’t) have had their eye on this one for some time.
Part of the reason for that is that this particular mutant has seemingly been on the rise. It was one of the founding viral types in Europe, and gained a good foothold there before being spread to other regions – and in those, the glycine mutant seems to have increased versus the original aspartate one. The tricky part is to distinguish between founder effects (what virus type got there first and started spreading) versus selective advantage (one that could show up later and outcompete what came before). In general, we don’t have quite detailed enough mutation-through-time data sets to be sure about distinguishing those two. The preprint applies several statistical tools to that purpose, and brings in another interesting piece of evidence from the Sheffield part of the collaboration: when you look at the “cycle threshold” for an RT-PCR test to pick up on the viral RNA, it looks like the glycine mutant needs a bit fewer cycles – which would mean that those patients the samples came from likely had a higher viral load to start with. Virologist Trevor Bedford at the University of Washington, whose name (and whose expertise) coronavirus aficionados will recognize immediately, has said on Twitter that the UW samples seem to show this effect as well. The preprint advances the idea that this mutant is more transmissible, and Bedford says that while there is some evidence for that, it’s not conclusive yet.
If, though, there is a difference, what could be behind it? The preprint notes that there are two broad possibilities. One is based on protein structure, and there are two main possibilities in that bin. The first is that that there’s an effect on the Spike’s receptor binding domain interaction with the human ACE2 protein. That’s harder to get a handle on, because this mutation is nowhere near the RBD. It would have to be some allosteric long range effect, and while that’s certainly possible, it’s difficult-to-impossible to try to model such things. You’re looking at a number of very small energetic changes lining up in the same direction to have an effect on a completely different region of the protein, and we just don’t have the ability to pick up on those computationally at the level of detail needed. The second structural possibility is that the interaction between the two main parts of the Spike protein (S1 and S2) has been affected by the loss of the polar Asp residue. It could well be forming a hydrogen bond with a nearby Threonine residue, which would disappear in the Glycine mutant form, and perhaps this “loosening up” facilities the viral entry process as the Spike goes about attacking the human cell membrane.
But the other broad possibility is immunological: the mutation is in a protein region that turned out to be important in antibody recognition during the original SARS epidemic. It showed high immunogenicity (many antibodies isolated from patients reacted to it) and it was also implicated in antibody-dependent enhancement. That effect has been mentioned here several times; it’s a major concern. Recall that this happens when antibodies bind to the virus but don’t neutralize it – in some of these cases, that antibody binding actually enhances the ability of the virus to get inside the human cell, which just makes everything worse. Not only do some SARS antibodies do that in the patients that developed them, but the same problem was seen in antibodies raised after potential SARS vaccine treatment. The good news is that everyone working on vaccines and monoclonals is very much on alert for this, and that so far more than one vaccine candidate has been reported to show no signs of it in animal models). So immunologically, the D614G mutant might be enabling antibody-dependent enhancement, possibly by favoring the generation of non-neutralizing antibodies to this region of the protein.
There’s even a mechanism that might bridge these two proposals: there’s more than one mechanism for ADE, and one of these (see the first link in the above paragraph) involves altering the Spike/ACE2 interaction. So you could have a direct effect on protein structure and target binding that also ends up with an immunological effect once the antibody response gets going.
It’s important to realize, though, that none of these mechanisms are proven. In fact, the very hypothesis that this mutant is more transmissible is as yet unproven – it’s not unlikely, but it could be wrong. Here’s an important question: what about the patients who are infected with with the D614G form? Turns out that the Sheffield group was able to match up hospitalization data with the sequencing of 453 individual patients, and they found no correlation of hospitalization status with the mutant form. Trevor Bedford and the UW group independently checked for this and had the same result: no apparent effect on severity of disease. At most, this mutation may be more transmissible (bad enough, to be sure), but it does not appear deadlier once a patient has been infected.
What about an effect on vaccine and monoclonal antibody development? We don’t have all the details, but the great majority of work I’ve been seeing on the monoclonals and on the antigen proteins targeted as vaccine candidates has focused on the RBD region of the Spike protein. This D614G mutation isn’t in that part of the protein, which is good. If that allosteric hypothesis has something to it, though, there could conceivably be an effect on the overall shape of the RBD, which could in turn affect antibody binding and selectivity. But this is two levels of un-proven-ness, at the very least. It’s also worth keeping in mind that that some of the monoclonal antibody candidates are from B cells of recovered patients, who themselves may well have been infected with the D614G mutant to start with. And the fact that there seems to be no difference in severity of disease between the two forms argues that the antibody response in general is unimpaired. So I’m not sounding any alarms on this based on the data we have.