So let’s talk about nanobodies – there’s a coronavirus connection to this, but it’s a good topic in general for several reasons. We begin at the beginning: what the heck is a “nanobody”?
The name is derived, rather loosely, from “antibody”. So let’s spend a minute on what antibodies actually look like. What you see at right is the three-dimensional structure of a typical one – you have a ridiculous number of these things circulating in your blood right now, nearly all of them subtly different from each other. The color codes are the two “heavy chains”, in red and blue, and the two “light chains”, in green and yellow. It all adds up to about 150 kilodaltons, a bit on the chunky side as proteins go.
It’s a bit easier to picture this stuff in schematic, so the next picture is general layout of these chains and domains. I’ve retained the same color scheme, but with some added information
There are the heavy chains and the light chains, as in the protein structure picture, but you’ll notice that the ends of both of those are variable regions, while the rest stay as a constant platform. Those variable ends (the “Fab”, for “fragment antigen binding” regions) are actually the part that recognizes antigens, as you would figure. You never know what sort of antigen you’re going to encounter next, and thus the insanely large collection of different antibodies that all of us are walking around with, produced “on spec” in hopes that one or another of them will happen to recognize what turns up. The two heavy chains are almost always identical in any given antibody, as are the two light chains (the antibody structure is symmetrical). All of this is held together by disulfide bonds between Cys amino acids and some other polar interactions between the constant regions.
I’m leaving out a lot. In any discussion of immunology that runs to less than about 500 pages in 6-point type you’ll be leaving out a lot. For example, antibody proteins have various sugar molecules attached to their surface at key points, and those are really important to their function. What’s more, there are actually five distinct types of heavy chain and two types of light chain – they’re mostly about the same size (a couple of the heavy ones are noticeably heavier) and they all fit into this same arrangement, but you can distinguish them by their amino acid sequences. That “base” at the bottom with the two heavy chains is called the Fc (“fragment, crystallizable”) region, and binds to various immune cells to regulate function. Meanwhile, the working details of the binding up in the variable Fab regions just get finer and grainer the closer you look at them; there are whole careers of work up there. Different classes of antibodies can also have these individual “Y” structures arranged further into pairs, for example, or into cycles with five of them in each unit. And so on.
Camelids and Nanobodies
And with that, we shall now abruptly veer off into talking about camels, llamas, alpacas and their kin, because they have their own variety of antibody. No one knew that they had a different system going until 1989, when a student-run project at the Vrije Universiteit Brussel was trying to come up with a diagnostic test to check camels for trypanosome infection. They discovered that camel antibodies were. . .weird. Some of them were just like the ones above, but about 75% of the camel antibodies (and up to 50% in the New World species like llamas) have no light chains at all. They just have the variable parts of the heavy chain stuck directly onto the “base” constant region. Sharks and their relatives, as it turns out, have something similar going on with a different sort of base region, in what are clearly two different evolutionary events: at least 220 million years ago for the cartilaginous fish and 25 million years ago for the camelids. Both sets of animals seem to work just fine with their proprietary systems – before these discoveries, most immunologists would have said that that such modifications would be likely to cripple the antibody response, but not so. That led to thoughts of clipping things down to just that heavy-chain variable chunk to see if those would recognize targets as well (i.e., having just one of those light-blue or light-red pieces in the above schematic by itself).
That they did, and “nanobodies” were born. Not only can they bind with high affinity to all sorts of antigen targets, but they do so via binding modes that have never been observed with real antibodies (which means that they might be able to recognize all those targets in new ways). They’re also much smaller than antibodies per se, which leads to some interesting properties. Nanobodies can have a wide range of stability and half-lives, which is tunable with some experimentation, and they often demonstrate much greater penetration into tissues and many other features. People have been investigating their properties and uses for over 25 years now. The Belgian researchers formed a company (Ablynx) in 2001 that has led the way, thanks to their solid patent positions in the area, but there have been so many twists and turns in the story that the first actual nanobody drugs have appeared just as their early patents have begun to expire (more on this timeline in that earlier link). With some irony in hindsight, some of those early investments in nanobodies were made in order to try to avoid the serious patent-licensing headaches with traditional antibodies.
There is now a preprint describing a screen for such nanobodies binding to the coronavirus. Update: the folks at VIB have also been working in this area, getting into it quickly, and I’ll be highlighting their efforts in another post. The team (a large multicenter effort led out of UCSF) had a yeast-displayed library of billions of potential heavy-chain fragments (prepared earlier in collaboration with the Kruse lab at Harvard), 21 of which ended up showing strong binding to the coronavirus Spike protein. These fell into two classes: Class I bound directly to the receptor-binding domain (RBD) and competed with the ACE2 from the surface of human cells. Class II, though, didn’t hit the RBD, but instead bound somewhere else and changed the conformation of the RBD so that it can’t recognize ACE2 when it’s available. When these nanobodies were put into an assay that measured the binding of fluorescent-labeled Spike protein and HEK293 cells expressing ACE2, the Class 1 species were active, but the Class II ones did nothing, weirdly.
Further work yielded cryo-EM structures for two of the best Class I candidates bound to the RBD, but they couldn’t get any such data for a Class II. That was worked out (partially) by another technique, where the complex was exposed to extremely reactive hydroxyl radicals (generated by synchroton X ray beams). You then look over the proteins to see what didn’t get eaten by the radicals. Those experiments showed a protected area on the Spike protein well away from the RBD, which is presumably where that particular Class II nanobody was binding.
The team went on to take one of the Class I candidates (designated Nb6) and make dimers and trimers of it, separated by inert Gly-Ser linking chains. The idea was that the Spike protein, which has a three-fold repeated architecture, could be inhibited even more strongly by binding to more than one RBD on its surface. And that proved to be the case: using surface plasmon resonance (SPR) assays, which let you follow on- and off-rates in detail, it became clear that the dimeric and trimeric forms of Nb6 occasionally bound with just one of its nanobody ends, in which case it could fall off again reasonably quickly, and also showed binding with all of its nanobody regions at once, in which case it came off much more slowly. The trimeric form showed subpicomolar affinity for the Spike protein in this assay, although the exact binding constant is so tight that it hasn’t even been quantified.
These various forms were taken into a pseudovirus cell infection assay (that’s where you rig up a harmless virus to use the coronavirus’ infection machinery, Spike and all). Plain Nb6 had an IC50 of 2 micromolar, and another Class I nanobody (Nb11) was almost the same. The best Class II nanobody (Nb3) was 3.9 micromolar. But that trimeric form of Nb6 (Nb6-tri) was 1.2 nanomolar in the assay, a two-thousand-fold-improvement. Trimer forms of Nb11 and Nb3 also improved, but not as much. In a test of Vero cell infection with real SARS-Cov-2 coronavirus (done at the Pasteur Institute in France), Nb6-tri prevented viral attack with an IC50 of 160 picomolar, which is truly impressive.
They didn’t stop there, though. A “saturation mutagenesis” experiment around the sequence of Nb6 was then tried, with new rounds of assays, and this yielded a mutant nanobody that was still more potent. You might wonder about trying to make things even better when you started out with billions of nanobody candidates in the first round, but a quick look at the math shows that a couple of billion nanobodies are just a speck compared to the total number of possibilities (around 110 amino acids, 20 variations per!) This one was trimerized as before, and the new mNb6-tri, when put into the SPR assay, showed no off-rate at all during the limits of the experiment, putting its binding constant somewhere in the femtomolar range at worst. It comes in with IC50s of 120 picomolar in the pseudovirus assay and about 50 picomolar in the wild-type infection assay, but those are probably at the limit of detection for both. Basically, we don’t actually know how potent this nanobody construct is, because we don’t have assays good enough to read out a number (!)
OK, now things get interesting. The authors tested the stability of mNb6-tri, and found that it can be heated, lyophilized (freeze-dried, basically), and nebulized into an aerosol with no loss of potency. It’s a very stable species that can put up with all sorts of handling and processing. You could certainly inject this material, just as you’d administer monocolonal antibodies. But there are more possibilities. How about formulating it as a nasal spray? Or in a nebulizer, to be breathed into the lungs, or even sprayed out into the room air? How about impregnating filter material with this protein so it pulls coronavirus particles out of the air as they pass through it? The extreme stability of nanobody proteins gives all of these a real shot, and they’re under serious consideration for development. The team says that they’re in discussion with several commercial partners to take this technology into human trials (and presumably medical-device trials, for the filtration idea), and I think that’s an excellent idea. This has real public-health potential, from the looks of it, and could be just the backup that we may need for the existing vaccine programs if they come in less effective than we’d like (or are rolled out more slowly than we’d like!) I hope that the money and resources are rounded up quickly.