HT-SPR Technology to Inform and Accelerate the Antibody Discovery Process

0:00:04.1 Speaker 1: I'm gonna talk today about how High Throughput SPR technology can inform and accelerate the antibody discovery process. So there are many approaches to biotherapeutic and biotherapeutic antibody discovery, and we feel that the value of Carterra's High Throughput SPR technology can be broadly applied to all of those approaches, and that it's highly efficient and uses very little sample and can be used on large numbers of samples early in the process. So we're gonna go through three recent examples which were published, of how High Throughput SPR can accelerate and add value to these antibody discovery workflows. So all modern biotherapeutic platforms can generate large numbers of clones, and these technologies have made huge leaps and bounds, but the competitive landscape is really growing, everybody has access to these technologies in general. I mean, there's obviously lots of diversity to them and different people have their specialties and core technologies, but the analytical tools need to adapt to properly access and address these modern scales of antibody discovery workflows and to really add value and richness upfront, to accelerate timelines and give more information early on to the scientists running these programs.

0:01:35.6 Speaker 1: So the conventional approaches of phage and hybridoma have worked well for many years and can generate large numbers of clones. Things are moving in the direction now more of B cell identification, or at least accessing B cell sequences, and the role of the ability to generate large quantities of synthetic DNA sequences really integrates into these workflows and some new independent ones using AI and more sophisticated phage, like with Twist and their Library of Libraries-type approaches. Also, advances in yeast and humanization technologies sort of build out the suite. All of these technologies are being majorly impacted by the implementation of next gen sequencing, where scientists can access the full genetic diversity in their sample sources in their libraries early on and use that to inform and evolve their workflows, leading to much more rapid and efficiency of discovery processes. So this is kind of changing the concept of the funnel, where instead of having a huge number of clones that have to go through very low resolution assays early on, you can assess your genetic diversity and make really strategic decisions about the type of data you generate early in the funnel, and how that genetic, or how that experimental information you get from these characterization assays relates to this whole collection of genetic diversity you have within your library.

0:03:23.1 S1: So if you have a good platform to generate large amounts of early characterization data and you have the next gen sequencing data, where you have good sequence information for all of those clones, and hopefully many more, you can feed that into bioinformatic systems and use things like machine learning and artificial intelligence tools to really give you a clear picture of how the genetic diversity in your set relates to the functional activity of clones within. And we believe that HT-SPR and epitope binning really serve as a core tool in this toolkit, or in this information area, because it allows you to overlay your genetic diversity, which you're getting from your sequence information, with the epitope diversity and the binding activity information from the kinetics. So you can really understand how sequence and function relate on a large scale that previously was unavailable. So I'd like to introduce the LSA platform. So this is an SPR system that uses array technology. It allows you to screen many more clones than conventional systems in a single experiment and in less time using only a small amount of the sample. This approach of High Throughput SPR is being broadly applied globally. Many institutions in addition to this have it, and it's used broadly across different antibody discovery workflows and approaches.

Array-Based SPR Platform

0:05:08.9 S1: So a real quick description of the system. So as I said, it's an array-based SPR platform, so it has two fluidic modules. One is a 96-channel side that can actually load or capture 96 samples at a single time and flow them back and forth across the surface to create your immobilized antibodies, and this can be done four times within an array to create a 384-spot array, then the system has a single flow cell which will dock over your created 384-spot array and flow one sample, so this makes the kinetics or the epitope binning highly parallel in nature, where you're collecting simultaneously data from one 270-microliter injection volume to the entire sample set that's captured. And these two systems work in an automated way, so they can come on and off the surface to run things like capture kinetics assays. So in addition to this array architecture of the platform, we have purpose-built analysis software packages that make processing of these large data sets quick and easy, but also visually rich, with dedicated tools for kinetics analysis and epitope binning. So I'm gonna go into some recent published examples of how High Throughput SPR is accelerating discovery workflows. So the first example I'm gonna touch on is some work done by Tom Yuan at Twist in collaboration with, actually Carterra in their Ebola glycoprotein project, and this was published in The Antibody Therapeutics in 2020.

0:07:00.1 S1: So this was an interesting project in that they had sequence information for, I believe it was over 200 anti-ebola glycoprotein clones and they were able, using their DNA synthesis platform technology, to rapidly create sequences and clone those into expression vectors for all of those clones. So about 14 days in, they had those constructs made and were able to express them in a high throughput recombinant expression platform, then in about two days, they did a first pass HT-SPR analysis on them with a focus on epitope binning. From that they were able to establish a broad set of communities and pick what they call pathfinder antibodies that they wanted to use to look deeper into the set and assess more clones. They also included, in this pathfinder phase, clones that bound to known structural epitopes on the ebola glycoprotein.

0:08:07.8 S1: So they re-expressed those clones and did another second pass High Throughput SPR experiment in another two days. And from that, they were able to get a really rich data set and so that's summarized well in the paper. You can see they have a competition map. So green means that the clones sandwich, red means they're competitive. For all of the clones, clustering them into these unique binding communities and when overlaid with these pathfinder antibodies and the known structural epitopes, they're actually able to create sort of a 3D model of their... Well, a 3D model existed of the ebola glycoprotein, but they're able to map onto that, the binding of all of these antibodies, all 233 clones. And with this information you can learn things about what mutations the antibodies are likely sensitive to based on the interaction phase of the molecule. You can also understand which confirmational state of the glycoprotein the antibodies are likely able to bind in and which ones can be combined together to make a cocktail.

0:09:25.2 S1: So this was a great early example of how a huge amount of structural information can be learned about a subset of clones or a set of clones in very little time but with really useful levels of resolution. Again, this was a really nice work by Tom Yuan at Twist, working also with Yasmina Abdiche at Carterra. Right. So we'll go into another example now of the COVID consortium, looking at the SARS-CoV-2. This is a large project with multiple players, and specifically, I'm gonna show data from LJI, La Jolla Institute for Immunology, Duke University and data I generated at Carterra. So the COVID consortium is the Coronavirus Immunotherapy Consortium and this was launched by the COVID-19 Therapeutics Accelerator, the Bill and Melinda Gates Foundation, Wellcome and MasterCard. And it's a global academic industry non-profit research collaboration from scientists around the world, so anybody who is making antibodies to COVID-19 was invited to submit them to LJI, who was organizing this, and they would allocate and distribute those samples to partner labs that would look at potency, efficacy, and LJI was doing a lot of structural analysis.

0:11:00.0 S1: All of this information goes into a database and then that enables LJI and the consortium to select likely antibody pairs that will make cocktails that will be highly efficacious and also highly resistant to mutations. So the LSA was used to do competition-based epitope binning. As of date, we've provided information to the consortium for 186 RBD binders, we have another hundred or so on the way, and the LSA was also used to determine binding kinetics of all of these maps to RBD, the COVID spike protein. This was done by the Tomaras lab at Duke University. They're determining the on-rate, the off-rate and the affinity values of those clones. This is from a recent paper from the consortium published in Science this month actually, well, in September, showing the kinetic data generated by Duke. This is obviously a very small subset. There were much more clones analyzed but these are some that are highlighted in the paper that bind two unique structural domains of the protein.

RBD Biding Kinetics

0:12:19.4 S1: You can see there's RBD binding kinetics on the left and full-length spike binding kinetics on the right. Some clones only bind to full-length spike 'cause they're not RBD binders, but in general you can see there's more rapid kinetics to the monomeric spike than the trimeric... Sorry, the monomeric RBD than the trimeric spike. But they're able to compare rapidly and with very little material the binding kinetics of all of these clones to various protein constructs, including mutants, and this was done by the Tomaras lab, especially Moses Sekaran and Jonathan Li.

0:13:06.4 S1: So this is some work I did at Carterra, where we took all of these RBD binders and did competition-based epitome binning, and this is the heat map generated. So competition is in red, sandwiching is in green. You can see there's a lot of complex overlaying competition profiles here, but also some very distinct epitope areas. So this can be summarized down into a set of communities. This is what we call a network plot, and it shows how these clones interact. Like I said, there's a whole lot of competition among these, where they're competitive with each other, but there's also pockets of very important and distinct differences. They are actually able to get a lot of resolution as to the uniqueness of these binding domains, even though there is a lot of competition within these profiles. Most of these clones are H2 blockers as they have been a bit pre-screened coming into this collaboration.

0:14:09.4 S1: This is the image of the dendrogram from the epitope binning software. So it takes the competition profiles of all these clones and figures out how closely they relate to each other. You can see at the top level we have the seven generalized communities, and then there's 15 sub-clusters, which actually when you drill down to that level, start making very discrete structural binding interface descriptions of the antibodies to the COVID RBD.

0:14:45.2 S1: This is a great figure that shows the cryo-EM or overlaid image of the epitopes from... Antibodies from these various communities. So LJI was doing this work Haoyang Li specifically, and I believe they have cryo-EM structures for about 70 of these clones. But what we can see is that the communities are highly descriptive of the specific interface on the antigen with the antibody. And so the ACE2 binding epitope is the dotted outline and then we have various communities. You can see this red RBD-1 almost exactly recapitulates ACE2 and there's, actually in the set, there's ACE2 SC constructs that are part of this bin, but we also have antibody clones that bind this type of epitope as well.

0:15:37.5 S1: Then the binding epitope, as the colored region starts to move around in the other communities, either shifting off the side or sliding down a face, or even interacting just at a different angle from this. And so using the sort of known, you could always think of these as pathfinder antibodies from the cryo-EM data, we can know exactly where all of the clones in the set bind and which face of the molecule. So again, this will inform the conformational state, the antibiotics can recognize their stoichiometry of binding and also what mutants of the receptor will be affected by the binding.

Viral Neutralization

0:16:17.7 S1: And this part addresses that directly. This is from Kate Hastie at LJI, where she did viral neutralization studies for many mutants of the receptor against these antibodies, and you can see that against the beta and the gamma, there are communities that are highly affected or very sensitive to these mutations in terms of losing activity.

0:16:42.2 S1: The Delta, it was a little different. There's still some pockets where it would be risky to take a clone forward from these bins because they're likely going to be influenced, but there's also other bins that appear to not be sensitive at all to these mutations and would be broadly neutralizing against known variants and mutants.

0:17:02.7 S1: Another view of this heat map here is showing the community's overlaid, but also labeling in the sensitivity to mutations from these bins, and fortunately, we have several bins that are discrete in that they don't have sensitivity to known mutations and they tend to sandwich with some of these others. So there's some opportunities here to make some really potent cocktails or very resistant-to-mutation antibodies.

Eli Lilly

0:17:37.5 S1: So the last example I'm gonna go into here is from Eli Lilly in the development of Bamlanivimab against SARS-CoV-2, and they actually were partners in this work with AbCellera. Both of these companies have LSAs and do this work.

0:17:54.4 S1: So in the beginning of January, as you all know, the world was caught a bit off-guard by the SARS-CoV-2 virus. The genome was published in January 10th, on January 10th, and then by August, the US alone had 5.4 million cases with 170,000 deaths. But in August, we had the first real approved working therapy when that was convalescent plasma, obviously not the most easy-to-access therapy. But in November 9th, only 11 months after the publication of the viral sequence, the FDA offered emergency authorization for Eli Lilly's Bamlanivimab, which is just a phenomenally short time for this type of therapeutic to be developed. Also, it beat the MRNA vaccines to market, which is impressive. They came on in December.

0:19:00.3 S1: And then shortly thereafter, Regeneron, shortly thereafter, later in November, Regeneron brought its antibody cocktail to market. This is just tremendous validation of the idea that biotherapeutic antibody discovery platforms are much more efficient than they have been in the past, but also one of the fastest ways to get a therapy into the hands of people and patients.

0:19:24.8 S1: So, Eli Lilly has published this discovery process with AbCellera in Science Translational Medicine in May of this year, and they go through their workflow and in about 23 days, they went from receiving the patient samples where they could isolate the antibody sequences to having a subset of 24 leads. And I wanted to highlight a phase where they spent about a week going from when they had 187 antibodies expressed to having full biophysical SPR-based characterization, so this is genetics epitope binning an ACE2 competition, and this was useful in making this down selection to 24 that went into later stage characterization and manufacturing ability and assessment. Then the SPR was heavily used in this rapid characterization phase with full kinetic profiling and full epitope binning establishing which communities these clones known sequence diversity went into, so they really understood the platform, I mean the importance of the genetic diversity on the epitope binding and the infinity of the set that allowed them to strategically down select into that limited number of clones.

0:20:50.3 S1: Also, another really interesting application of the LSA in this process was, this is work done by Denise and Foster, they were able to take 96 mutants of the SARS-CoV-2 RBD, captured onto the array, and using the Bamlanivimab Fab, do full kinetics and binding to that. So this allowed them in basically a day, once they had the constructs made, to do this rapid characterization of full affinity binding kinetics to all of the mutations. So this is a quote from that same paper that it was really a testament to the characterization and discovery platforms that they've developed and some existing collaborations that they were able to go from starting the process to into a patient in 90 days. It's just amazing speed.

0:21:46.2 S1: So in conclusion, High Throughput SPR enables a richness of kinetic and epitope data which was previously unavailable, and especially being able to access this information early, it has broad applications across all types of clonal sources in antibody discovery workflows. Next generation antibody workflows with high clone number and full sequence information may benefit the most of all, as the sequence diversity within the set can be rapidly and meaningfully understood and characterized, and the ability to access this information early in the funnel and in a very complete way really can guide a better clone selection and accelerate these antibody discovery workflows. Right. Well, thank you for your time.