In this edition of Carterra’s Scientist to Scientist Interview, luminary scientist and leader Andrew Bradbury PhD, Chief Scientific Officer of Specifica, discusses how they discovered SARS-CoV-2 neutralizing antibodies in just weeks. Daniel Bedinger PhD, Applications Science Team Lead from Carterra and Andrew also discuss next generation antibody library platforms and how HT-SPR technology powers their work.
Posted by Daniel Bedinger, PhD
0:00:00.0 Dan Bettinger: Hi I'm Dan Bettinger, Application Science Team Lead at Carterra and I'm here with Andrew Bradbury, who's the Chief Scientific Officer at Specifica. Andrew, thank you for speaking with me today.
0:00:14.3 Andrew Bradbury: Thanks, my pleasure, Dan.
0:00:16.4 DB: Yeah. Great. So let's share some slides. So Andrew, can you tell me a bit about Specifica's antibody discovery platform?
0:00:25.2 AB: Yes, so this is something I've always wanted to do. I was unable to do it in an academic environment, people weren't really interested in funding it, but within the commercial environment, it's a different ballgame. And so, what we tried to do was to establish a discovery platform which was going to really provide excellent anti-bodies and in doing that what we decided to do was to start off using well-behaved clinical antibodies as scaffolds. The idea being that if an antibody had been into patients and behaved well, then that would be a good scaffold to use for further diversity. Then within that scaffold to insert diversity, which is substantially liability-free. So what do I mean by liabilities? Well, we have a lot of next gen sequencing in Specifica, and we're able to identify different CDRs and eliminate sequence liabilities like glycosylation sites, deamidation sites, isomerization and the like. The result of this is high quality CDRs going into high quality scaffolds, which result in a high functional diversity.
0:01:35.7 AB: So if your sequence it, we have over 90% of the antibodies are open reading frames, but because of the way the library has been designed, we also have a very high proportion of well-folded antibodies as well. And then we build libraries for people. So every library is exclusive. We use donors only once. We never use them again. The idea being that if you select an antibody from your Specifica library, you can be confident that nobody else will be selecting the same antibodies. This has been a problem where the same target has been selected against the same... From the same library by different companies. And so the net result of all this is that over 80% of our tested antibodies have no biophysical liabilities. So the difference in a biophysical and sequence liability is that the biophysical liability is something you can measure so it has no aggregation, no polyreactivity and so on. Over 60% of tested antibodies have affinities better than 10 nanomolar, with 20% of them subnanomolar. And usually when we do a selection, depending upon the target concentration we use, we get between 500 and 5000 different clonotypes per target, where each clonotype differs from a different clonotype by an average of 20-40 amino acids in all the CDRs.
0:02:54.3 DB: Wow, that's a lot of diversity. So I'm very interested in the thinking that goes into designing antibody screening and selection processes. So can you comment what you feel are the most important considerations for early screening and selection?
0:03:12.8 AB: Right, so I think the most important is to get a very broad paratopic or sequence diversity. And the reason for that is that not all antibodies will have biological activity, and you want antibodies binding to as many different epitopes as possible in order to identify those antibodies that bind to an epitope, which is likely to give you biological activity of interest. So the way that people have done this traditionally is just to screen more and more antibodies, more and more clones, but even if you get up to 10,000 different clones or... It's usually not 10,000 different clones, it's usually 10,000 clones, many of which are repeated. I would argue that's a low throughput approach. So what we do is we combined picking with next gen sequencing and this allows us to... In the picking, to identify antibodies that we have found within the next gen sequencing. And within the next gen sequencing, we're able to identify different antibody sequences that are worth synthesizing using gene synthesis.
0:04:25.7 DB: That's very cool. So you're ensuring good sequence diversity before you go into the epitopic characterization and kinetic characterization?
0:04:35.4 AB: Well, that's what we like to do, is we like to get as broad a sequence diversity as possible, and then test those antibodies for binding and do things like binning afterwards. Now, if all you're doing is testing picked antibodies, then you may have a lot of sequence duplication. Such sequence duplication may correspond to completely identical antibodies, or alternatively, it might be antibodies that belong to the same clonotype, so even though they have slightly different sequences, they would be expected to have the same biological activity albeit at different affinity levels.
0:05:18.3 DB: Great, well, that's a good lead into my next question, which is, can you talk about how high throughput kinetics fits into your workflow?
0:05:26.7 AB: Right, so... Measuring affinity has always used to be something that was quite painful until the SPR, the LSA that Carterra's developed came along. And so, using this, we're able to easily get 96 affinities relatively straightforwardly, and we can increase that as well. And the affinities that we're getting are really quite remarkable from the point of view of the antibodies as well as the data. So if you look at the data that's being shown here, you can see that the curve fits, which is red compared to the experimental, which is in blue, you can see those fits are on the whole really very good, indeed. And so we're very confident in the affinities that we're getting out using the LSA.
0:06:09.4 DB: Yeah, that's great. Quite a few high affinity binders as well. So can you tell us a little bit about Specifica's efforts to generate antibodies against the SARS-CoV-2 spike protein?
0:06:23.3 AB: Right. We, being an antibody company when SARS started, we sort of stood on the sidelines for a bit, but after a while it really became impossible not to try and do something. It's like a moral obligation actually, and so we got hold of various forms of the spike protein, we made antibody libraries from patients that had extremely high titers, which are illustrated in this slide, so the immune kappa and the immune lambda are two libraries that we made using V kappa or V lambda with a heavy chain output from a patient had a very, very high titer. And the plots that I'm showing there, in the x-axis, you have the amount of antibody displayed on each individual yeast, each dot is a yeast, and in the y-axis you have the amount of binding to the S protein. And so what you're interested in is yeast that are found in the upper right quadrant, because those yeast in the upper right quadrant are both displaying large amounts of antibody and also binding large amounts of target, which in this case is the spike protein.
0:07:30.0 AB: And what you can see that one nanomolar concentration of spike, the pattern of the Gen3 library, which is a completely naive library, which is the one on the top, top row, is very similar to the immune kappa and better than immune lambda. What does this mean? This means that the antibodies that we're able to select using the Gen3 are essentially very similar to the antibodies that you can get from an immune library, so we believe that you can now get antibodies as good as immunization without having to immunize.
0:08:05.3 DB: Wow. That's a really powerful demonstration. Thank you. So looking at some of the kinetic information you've shared about these antibodies, I guess maybe you can speak to this a little bit and are there specific kinetic criteria you use in your selection before you push clones further along in the funnel?
0:08:30.1 AB: So what we tend to do is because we combine phage plus yeast display, we tend to... It all depends actually on what the customer wants for their antibodies, not everybody wants high affinity antibodies. So for example, with CAR T-cells, you want single chains, it's not obvious that high affinity is necessarily better than low affinity. And so we always discuss with partners what their requirements are before designing a selection strategy. In the case of SARS-CoV-2, we thought that yeah, high affinity is gonna be the best because it's likely to be the most neutralizing and so what we're showing here on the left hand side is affinities, which is sorted by the affinities to the trimer in orange, but also are showing superimposed and that's the spike primer, also showed superimposed the binding to the receptor binding domain of the spike protein. What you can see is there's actually little, not that much of a correlation between the trimer binding and the RBD binding. The best affinities that we were able to identify recognizing the RBD, that's a monomeric binding, that was 15 picomolar, and the best, and we got a similar affinity for the best against the trimer, which is also a 15 picomolar, as you can see there.
0:09:42.3 AB: Now, what I was saying earlier about antibodies coming from the libraries, our naive libraries, are as good as those from immune libraries, is illustrated on the right-hand panel, where in orange, there are some antibodies that Dennis Burton sent us. These were isolated from convalescent patients, and you can see that the iso-affinity plot has 10 picomolar, 100 picomolar and so on. And you can see that all those orange dots are well within the pattern of blue dots, in fact, if anything, the blue dots may even be better than the orange dots. And so, in terms of the affinity here, we are able to get antibodies that are again, confirming as good as those coming from immunization. What we have found, interestingly, is that, at least in the case of these antibodies that we've pulled out, we don't show a great correlation between the affinity of an antibody for either the spike trimer or the receptor binding domain for the IC50 of the antibody.
0:10:41.8 DB: Well, that's a good point. I mean epitope is foundationally important for potency and MOA of the antibodies, so can you tell us a little bit about what you learned about your antibodies by doing epitope binning or how this aided your characterization?
0:11:00.2 AB: Yup, so that's another nice thing about the LSA is that you can bin antibodies relatively easily. So what we did here was we tried to bin the antibodies we selected against those that had been previously identified by Dennis Burton. And as you can see here, what we found was that most of the antibodies... Well, in fact, pretty much all the antibodies we pulled out, binned into what we... This bin we called bin 3B, and that included the two script antibodies as illustrated. In addition, there was another bin called 4B, which was a couple of antibodies... Sorry, bin 4B had two antibodies from the scripts, and we were able to separate the binding of those two antibodies from some of our antibodies as well, so the great thing about doing binning on the LSA is that you can really go deep into the binning and understand where antibodies are binding at the same site and where they're binding differently.
0:12:02.8 DB: Right, yeah, the more antibodies you include in these binning maps, the higher resolution sort of the picture becomes, which I think is one of the neat differentiators of the LSA over lower clone number inclusion binning assay 'cause you get this finer resolution and differentiation.
0:12:20.8 AB: So this is the sterile, this is the neutralization data I was talking about.
0:12:29.0 DB: Yes. So obviously congratulations, these look like very potent antibodies. Happy to hear anything else you'd like to say about this data or these antibodies, but also what are your plans for these antibodies going forward, or can you even say?
0:12:46.4 AB: That's a great question. So the antibodies... What we have here... This is data from IC50 on pseudo-viral inhibition and again, you can see that all the antibodies we pull out from our naive library are in blue, the two scripts antibodies are in red, and the IC50s we're getting here are all in the sub 10 nanograms per ml range with some being less than two, as is indicated there. And if you look to the right, you look at the affinities for the RBD, you see there's actually very little correlation, as I mentioned earlier, between the affinity and the IC50. We now have got live virus neutralization titers for these antibodies, and essentially they map very similarly to the pseudo-viral neutralization titers. We have one antibody that is down to 1.3 nanograms per ml. Although we haven't repeated that so it could be that that will move around a bit. But these are really ultra potent antibodies up there with most of the best that are out there, with perhaps the exception of some of the Regeneron antibodies.
0:13:55.7 AB: Now, what are we gonna do with them? We arrived late to the game, as I said. It seemed to us a few weeks ago that these were going to be an interesting set of antibodies. We still don't know if there's any commercial interest in them, but we're certainly happy to think about commercializing these antibodies. We're in the process of testing sequences and escape mutants and interestingly, these antibodies, at least so far, appear to be quite resistant to escape mutants. We're trying to understand what exactly that means, and in collaboration with Ian Wilson in the scripts, we're looking at the structures of some of these antibodies. And a lot of the pseudo-viral and the live-viral neutralization studies have been done with Abe Pinter's Lab in Rutgers and also Dennis Burton in the scripts. So we're open to commercializing them, but we're also continuing to work on them just because we think that they really demonstrate the power of this naive antibody library platform.
0:15:00.6 DB: That's great. I'm excited to hear that people have stockpiles of potential escape mutant, neutralizing antibodies in the fridge in case they're needed in the future.
0:15:13.9 AB: Right.
0:15:15.1 DB: So, you can take us out by highlighting any benefits or any closing comments on Specifica's generation 3 antibody discovery platform.
0:15:26.1 AB: Yeah, thanks, thanks, Dan. Well, I think ultimately, basically, what we're getting is we're able to get drug-like antibodies directly from the libraries, the platform, without any further downstream affinity or specificity or other sorts of maturation. So as I mentioned earlier, over 80% of tested antibodies have no biophysical liabilities and the rest we found have only a single liability. Over 60% of the antibodies have high affinities with 20% subnanomolar affinities. Very, very high antibody diversity, 500-5000 different clonotypes. And we found that if you do next gen sequencing, as we've done, you can actually expand the epitope space that you're exploring by between two and maybe 10 or 20 times, depending upon how many antibodies you're picking, if you're just doing the low-throughput approach, and then finally the antibodies were pulled out of advanced SARS-CoV-2 are really antibodies that are pretty much as good as, and in most cases, better than most immune antibodies. So I think that we finally managed to solve the problem of how naive libraries can compete with immunization.
0:16:43.6 DB: Certainly it seems that way.
0:16:46.5 AB: Maybe I should say, just before we finish. At Specifica, we do antibody discovery. So we're happy to do discovery if people are interested, but as I also said, we also provide the whole platform to partners and this includes the a phage library, very high diversity, the ability to move over to yeast protocols, training, the whole works, plasmids. So the idea is to really set people up to be able to go out and discover antibodies they want.
0:17:22.5 DB: Great, well, Andrew, I really wanna thank you for joining me today and describing your antibody discovery platform, and we wish you great success with your company's development of these antibody generation platforms moving forward. So thank you for your time today, we really appreciate it.
0:17:39.9 AB: Thanks so much for taking the time to interview me. It's been my pleasure