There’s a lot of work being done on antibodies for the coronavirus and on the protein domains they recognize. This of course has bearing both on the idea of monoclonal antibody therapies and for the vaccines that are in development, so let’s have a look at the new data. For reference, here’s a background post on some of the proteins that the virus makes and the mutations that have been spotted in them, here’s my post on the basics of antibodies and immunology as relating to the epidemic, with an update here, and here’s my earlier post on monoclonal antibodies as a treatment, now updated with some new items.
So you’ll see from that last link that what we’re after are neutralizing antibodies, not ones that just bind to the coronavirus without interfering with it enough to be useful. The particular protein that everyone believes is the natural target for such things is the famous Spike, because it’s well-exposed on the surface of the viral particle and is crucial for the virus to enter human cells. In particular, the receptor-binding domain (RBD) of the Spike is the part that interacts with human ACE2 to start the viral entry process, so that’s had a huge amount of work directed toward it, much of which is very much informed by the work done on SARS and MERS (here’s a new review). In fact, some of the antibodies developed for those have been shown to cross-react with the current viral spike protein, although this is by no means a universal property.
We now also have several reports of antibodies that have been raised to such proteins – some of them isolated directly from patients, and other monoclonal ones made by techniques that are more directly amenable to protein scale-up. Obviously, the companies working in this area have such antibodies already in development, but we don’t know much about them yet. Here’s a report from a team in China that isolated monoclonal antibodies by finding memory B cells from recovered coronavirus patient blood samples. Their three best candidates all bound to the Spike protein’s receptor binding domain, but only two of them interfered with the association with human ACE2.
That might make you wonder about the third one, but here’s a new open-access paper from a Dutch team (Univ. Utrecht and Erasmus/Rotterdam) that describes an antibody like that in detail (a preprint on some of this came out back in March). They produced spike proteins from several coronavirus pathogens (the latest one, SARS, MERS) and immunized mice with them sequentially over two weeks. These weren’t just any mice, naturally – they were H2L2 animals, a trademark of Harbour Antibodies for a strain that has a partially humanized immune system. B cells from the mice (spleen and lymph) were fused with myeloma cells in the classic “hybridoma” technique that produces new immortal antibody-producing cell lines. They produced 51 different lines, and screened those to find four of them that cross-reacted with the current pathogen’s spike protein. One of those (named 47D11) neutralized viral infection of Vero cells in culture, and that one is the focus of the paper. (The chimeric antibody was modified to be fully human as it became of interest for further experiments).
This antibody does indeed bind the RBD of the Spike protein, and it binds to both the SARS coronavirus and to the current one (although it’s even more potent for the earlier variety). But there was a surprise: assays (biolayer interferometry, for the specialists) showed that the antibody did not keep the Spike protein from binding to the ACE2 protein. It did, though, stop the next stage of the viral entry process (syncytia formation, where the outer membranes of the virus and the human cell start to merge), so there’s an unknown mechanism at work here that doesn’t involve direct competition with ACE2 binding. That’s something to keep in mind – there are a lot of steps involve in viral infection, and several ways to interrupt it with an antibody. A paper last year on antibody modes of action against SARS and MERS may have some insights on how such a mechanism might work, if people want to dig into the details. But it seems clear that failing to knock down the RBD/ACE2 interaction alone is not enough to disqualify an antibody candidate.
Now, even that RBD region has subregions. It’s 168 amino acids long, and that’s a pretty good amount of real estate. There’s a core domain that contains a smaller 60-residue exposed subdomain that loops out – that’s the part that first recognizes the ACE2 human protein. That is the tip of the spear: the receptor-binding loop region poking out of the receptor-binding domain of the Spike protein of the coronavirus. The problem is, if you target that particular region (which at first sounds like something you’d want to do) you run the risk of losing activity due to mutations, because that region is one of the more variable ones in the whole Spike. This 47D11 antibody, though, binds to the core domain of the RBD, which is much less variable, and that core-domain binding mode fits in with the way that it doesn’t seem to interfere with the ACE2 recognition event. And as the authors mention, this opens up the possibility of combining this antibody with another one that binds the exposed loop as a dual-acting therapy.
Meanwhile, the Israel Institute for Biomedical Research announced, without any real detail, that they also have a neutralizing antibody. The word “breakthrough” was thrown around by Israeli officials, but I wasn’t as impressed. At this point, finding a neutralizing antibody is not really news, because most recovered patients (not all!) have recovered by generating them. What’s of interest are the details past that: where does this antibody bind, exactly, and with what affinity? What part of the viral life cycle does it disrupt? The paper from the Netherlands goes into those details (without ever actually mentioning the protein sequence of their antibody, though, as many have noticed). But the Israeli announcement is best met with a shrug until more details are released. We can expect more announcements of this type in the near future, I would think.