I’ve had several questions about antibody-dependent enhancement, which has always been a worry as the coronavirus vaccines have been developed. I figured it might be worth a look at just what we know about it, why one might be worried, and why (on the other hand) one might be hopeful.
The simple definition of ADE is “raising antibodies that don’t protect, but actually make a viral infection even worse”. And obviously, that’s the opposite of what you want. Remember that there are “neutralizing” antibodies as opposed to non-neutralizing ones – a neutralizing antibody, as the name implies, binds to its target in a way that shuts its function down. That’s generally done by blocking the “business end” of a given protein target, smothering the binding surface that it would need to do its usual job. For the coronavirus, a straightforward example of a neutralizing antibody would be on that binds to the tip of the Spike protein, the receptor-binding domain (RBD) that is the part that recognizes and binds to the human ACE2 protein on a cell surface. Block that thoroughly enough, and it would seem that you have blocked the virus’s ability to infect your cells.
There are other ways, as this blog post earlier this year makes clear (or tries to!) You don’t have to just completely cap the end of the Spike protein to shut it down – as it turns out, you can bind an antibody further down the Spike and have it be neutralizing, just so long as it interferes with the structural changes that the Spike protein needs to undergo when it starts binding to a human cell membrane. There’s a whole subunit of the Spike (S2) that is involved in the membrane-fusion step, so throwing a wrench into that will work for you, too. Proteins make all sorts of adjustments as they fit to each other: this part has to shift down and over, that bond has to rotate, and some of those adjustments may be non-negotiable (and thus targets for blocking the whole process).
So there are plenty of ways to get neutralizing antibodies, with various ways of binding to the Spike protein and in binding to other coronavirus proteins as well. But there are also plenty of ways to get non-neutralizing antibodies, ones that stick to some part of the coronavirus particle without really inconveniencing it much. That’s obviously useless, and antibody-dependent enhancement takes things down another notch, from useless to outright harmful. With ADE, the binding of such an antibody actually assists the virus, by (for example) actually making it easier for the virus to get taken up through the outer membranes of human cells. Another possibility is that an antibody that would be neutralizing if present in sufficient amounts can actually enhance infection in lower dilutions, which has been seen with influenza antibodies and other viruses as well. This seems to be through aggregation of viral particles, although other factors might be at work.
You really don’t want ADE through any of these mechanisms – bad things happen. Dengue fever is a classic example, because it infects humans through four distinct serotypes. If you are infected with one of these and raise a successful immune response, you may well be at increased risk of serious infection with one of the other serotypes. The neutralizing antibodies for one of the types are often not neutralizing for the others, but instead allow that cell-antibody-receptor mechanism to kick in (easier infection of human monocytes), known as “extrinsic ADE”. There’s also an “intrinsic ADE” seen with dengue, which leads to greater viral replication inside infected monocyte cells before they burst and release their contents. The mechanisms for that are still being worked out, but seem to involve suppression of cytokine pathways.
ADE has been seen with HIV infection (where it may be mediated by one of the complement pathways, which kicks in after an antibody binds to its target), with Ebola (where a completely different complement-driven mechanism seems to be operating), with coxsackievirus (and other picornaviruses), and in many others. It should be noted that inappropriate complement activation can cause troubles of its own, which can contribute to the severity of ADE-driven disease – this is particularly noticeable in respiratory viruses (influenza and others) and their effects in the lungs.
Evolutionarily, you’d figure that developing such things would be under positive selection pressure: higher organisms are constantly fighting off viral infections by raising antibodies to them, so something that causes this to backfire would probably be an advantage for any virus that hits on such a mechanism. So ADE is not some weirdo exception in viral infections, unfortunately – it’s pretty widespread.
And in the same way that viral infections can involve ADE, so can the antibody responses raised by vaccines. There was an inactivated-virus vaccine tried in the 1960s against RSV (respiratory syncitial virus) that in human trials actually caused infants to come down with worse cases of RSV. This effect has been duplicated with RSV in cell cultures and in primate models, and one hypothesis has been that (as with the extrinsic-ADE of dengue), the exposed regions of the antibodies bound to the viral particles bind in turn to receptors on human cell surfaces, and allow them to be taken into the cell more directly. A 1960s inactivated measles vaccine candidate showed similar effects.
Here’s a recent paper taking all this into the context of the current pandemic. And since this post up until now has been rather gloomy, you’ll be glad to hear that the news starts to improve at this point. For one thing, the current coronavirus does not seem to productively infect macrophages, which are by far the main target for that antibody-receptor-uptake ADE mechanism. The related MERS coronavirus was able to do this, but not SARS-CoV-2, fortunately. So the two mechanisms seen in (for example) dengue do not seem to be as much of a worry. The complement-driven stuff is still on the table, though, and indeed matches up well with the “cytokine storm” lung damage seen in severe patients.
But as that new paper says, thus far “No definitive role for ADE in human coronavirus diseases has been established.” That may be a bit surprising, if you’ve been seeing worried stories about antibody-dependent enhancement over the last few months. That doesn’t mean that ADE can’t be operating, of course, just that we don’t have the solid evidence that it is. Another surprise in that line: there’s been a lot of talk about a possible protective effect of prior infection with the other respiratory coronaviruses. Well, there’s a flip side: antibodies raised against those could potentially make SARS-CoV-2 infection worse through ADE, if they’re non-neutralizing.
Now, there have been worries about ADE with coronavirus vaccines as well. This is another case where having all the work done against the 2003 SARS epidemic has paid big dividends this year. Some of the earlier attempts at a SARS vaccine showed ADE effects in mouse models, and further work showed that this seemed to be linked not so much to the antibody response as to the T cell response. Specifically, a “Th2” heavy response (as opposed to more Th1 or a balance between the two), was linked to lung pathology. Those are subdivisions of the CD4+ T cells, based on which cytokines they produce, and these results alerted everyone to keep an eye out for that. Mouse immunogenicity studies with the current vaccine candidates did not show these effects.
In primate models, there were reports on the earlier SARS front like this one: four different peptides as vaccine candidates, three of which seemed to generate protective responses and one of which made things worse. But that also reminded everyone to watch carefully, and it has to be noted that the primate models for the current SARS-CoV-2 vaccines showed no signs of this, either. Not all the earlier SARS work in primates did, either: these two studies went well, with no ADE signs. But immunology being what it is, one has to watch carefully as you move into humans, and the clinical trials that we have been seeing read out have been alert to these possibilities. So far, so good.
This has been why we’ve seen so many vaccines taking care to put the Spike protein into its “prefusion” conformation. The worry has been that if antibodies are generated to it after it’s had a chance to bind to human cells, that gives you a better chance for nonneutralizing ones (and thus potentially a better chance for ADE). And you’ll have noticed the emphasis on neutralizing antibody titers along the way as well – that would have been there anyway, but a high proportion of outright neutralizing antibodies is also a safeguard against antibody-driven enhancement of disease.
At this point, I would say that the main worry for any ADE effects would be if the coronavirus mutates to the point that the antibodies generated by the current vaccines become non-neutralizing. And honestly, I don’t see that happening (it certainly doesn’t seem to have happened yet). Targeting the Spike protein is another big benefit that we got from the earlier SARS work; which suggested that (for example) targeting the Nucleocapsid (N) protein was riskier. With the Spike, you put the virus in an evolutionary tight spot: evading the antibodies while trying not to lose the ability to bind to the human ACE2 protein. So far, that looks like too narrow a path for the virus to stumble through.