What is Therapeutic Antibody Discovery and Why is it Important?

The meteoric rise in biotherapeutic antibody production is directly linked to their clinical success in treating a wide range of diseases and providing transformative medicines to patients. Evaluation of potentially therapeutic antibodies has historically been a slow and tedious task consuming large amounts of researchers’ valuable time. However, in recent years, high-throughput surface plasmon resonance (SPR) has streamlined the way investigators approach their drug discovery workflows. The ability to characterize binding kinetics, affinity, and epitope specificity on large antibody panels with minimal sample consumption at early stage research is accelerating the library-to-lead triage. In this GEN webinar, you will learn how high throughput SPR can be used to rapidly generate high-quality kinetic data from 384 antibodies in parallel from crude material and with minimal sample requirements. Additionally, we will demonstrate how epitope binning assays can be performed routinely on up to 384 antibodies per array, facilitating the identification of clones targeting unique epitopes.

Kathryn Ching, PhD
Senior Scientist
Ligand Pharmaceuticals

Yasmina Noubia Abdiche, PhD
Chief Scientific Officer


0:00:00.0 Kevin Davis: Hello everyone and welcome to this special webinar, produced and hosted by Genetic Engineering and Biotechnology News. This webinar is sponsored by Carterra and is entitled Accelerating the Discovery of Therapeutic Antibodies using High Throughput SPR. I'm your host, Kevin Davis, the Editor-at-large with Genetic Engineering News. A quick summary of what we're going to be hearing today, evaluation of potentially therapeutic antibodies has historically been a tedious process. But in recent years, high throughput surface plasmon resonance or SPR has streamlined the way investigators approach their drug discovery workflows. This ability to characterize binding kinetics affinity and epitope-specificity on large antibody panels is accelerating the library to lead triage process.

  • How High Throughput SPR Can Rapidly Generate High Quality Multiplex Kinetic Data?

0:00:57.6 KD: Today, you'll hear how High Throughput SPR can rapidly generate high quality multiplex kinetic data with minimal sample requirements. We'll also hear how epitope binning assays can be performed on hundreds of antibodies per array facilitating the identification of clones targeting unique epitopes. So, let's go ahead and meet our speakers. We have two great speakers today, first, you'll be hearing from Kathryn Ching, who is senior scientist at Ligand Pharmaceuticals, and we'll also be hearing from Yasmina Noubia Abdiche, the Chief Scientific Officer at Carterra. Following both presentations, we will have an interactive Q&A session during which we will try to answer as many of your questions as we can. So to participate, simply hit the Ask a question button below the presentation screen, type in your question and hit submit. Alright I think we'll get our webinar underway without any further ado, Kathryn, a warm welcome to you and over to you.

  •  Katie Ching - Scientist at Ligand Pharmaceuticals

0:02:04.6 Kathryn Ching: Good morning. My name is Katie Ching, I'm a scientist at Ligand Pharmaceuticals. And I wanna first start by thanking Yasmina and the team at Carterra for inviting me here this morning to participate in this webinar. We've been working with Yasmina and her team for almost five years, and really without their help and without their technology, we wouldn't know as much as we do about the antibody repertoire of our OmniChickens. And I'll be talking to you a little bit about our newest genotype today, and the work that we've been doing with Carterra to characterize that. You've probably heard of Ligand through the OmniAb platform for antibody discovery. There are three different species that are part of the platform, the OmniRat, the OmniMouse and the OmniChicken for traditional heavy light chain antibody discovery. But we also have two animals, the OmniFlic which is a rat and the OmniClic for bispecific antibody discovery. As part of the chicken team at Ligand, one of the first questions I always get asked is, why chicken? 

0:03:16.5 KC: And the answer to that is evolutionary distance. So, we are about a 100 million years distant from a mouse and about 300 million years distant from a chicken. So what that means, and it's sort of illustrated down here in a cartoon format, is that when we're looking at really highly conserved mammalian proteins, there's a significant amount of shared homology between a murine and human ortholog of a given protein. And that means that there's a lot less epitope space on that protein that the mouse sees as foreign, and that's gonna be reflected in your antibody repertoire. Now, in contrast, if we take that same protein and we immunize a chicken with it, because of that increased genetic drift, that distance of relatedness, there's a lot more on that protein that the chicken sees as foreign. And again, this will be reflected in the antibody diversity that's generated upon immunization. So, a greater evolutionary distance means a protein is gonna be more immunogenic in the chicken, and that will be reflected by greater antibody diversity, greater epitope coverage and probably differences in kinetics.

  • OmniChicken

0:04:38.5 KC: So, in the OmniChicken, we've knocked out the chicken heavy and light chain loci and we've replaced them with human B regions. So, I should mention that the chicken immune system is a little bit unique. They have a single V gene at their heavy and light chain loci. And they undergo VDJ recombination, but they don't actually generate a lot of diversity from that process. They generate diversity through a process called gene conversion, in which upstream pseudo-genes which are promoter less and don't contain RSS sequences, donate stretches of sequence to the rearranged V, and then chickens also undergo thematic hyper-mutation in the periphery. So what you're looking at here are the two different heavy chains that we have available in our OmniChickens. The first we call SynVH-C, and that's a pre-rearranged VH-323 downstream of human pseudogenes derived from naturally occurring human ESTs. And then our other heavy than we call SynVH-SD, and that's actually a rearranging heavy chain again, VH-323. But there we're looking to get diversity and sequence line at CDRH-3 and both of those transgenes are upstream of the chicken constant region.

0:06:08.5 KC: On the light chain, we again have two different transgenes available. SynVK-CK is a pre-rearranged V-kappa 3-15, again downstream of an array of human pseudogenes, and then we have what we call SynVL for V lambda1-44 also pre-rearranged, and that's what I'll be focusing on today. Besides being kappa versus lambda, our kappa is upstream of a chicken constant region, and our lambda is upstream of a human constant region. We basically have four different types of birds available for antibody discovery, a pre-rearranged VH3-23 and a re-arranging VH3-23, each paired with either the kappa or lambda light chain. I just wanna quickly show you that when we look at the B-cells in the periphery of these birds, we see that they have normal B-cell populations. So, on the top here you have wild type bird, in the middle, heavy chain knockout bird, and on the bottom one of our OmniChickens, this is actually V kappa with a pre-rearranged VH3-23. And you see cell surface staining, Bu-1 is a marker for chicken B-cells, you can see a nice bright population here in the wild type birds absent in the heavy chain knock out 'cause there is no IgE at the surface.

  •  Population Recapitulated in the OmniChicken

0:07:42.0 KC: And again, that population recapitulated in the OmniChicken. The same when we look at heavy and light chain staining in these birds, and we see it very nicely in the wild type birds, absent in the knockout, and again present in the OmniChicken. And on the right side of this slide, you're just looking at some double staining, showing that these are indeed B-cells staining with IgM or IgL or on the far right IgM and IgL together. As I mentioned, one of the great advantages of the chicken is that we can generate antibodies to highly conserved targets. Human BDNF is a project that we've been working on, human brain-derived neurotrophic factor is about 97% conserved between humans and mice, and actually about 91% conserved between human and chicken. The difference being, it is incredibly difficult to get a titer when you immunize the mouse with this protein, but as you can see here, these are examples of four different birds with four different combinations of heavy and light chain. Excuse me. We see a very strong titer on the first draw from all four of these birds, whether or not it's carrying a pre-rearranged or rearranging heavy chain or kappa versus lambda.

0:09:08.0 KC: And we've generated a panel of about 100 unique antibodies to this bird that we're working on characterizing now. One of our main focus is... And here is where we've worked really closely with Carterra, is characterizing each of our new different lines of birds against the same protein. And so this gives us a really fairly complete idea of how these different transgenes compare, 'cause we're looking at them, immunized against the same antigen. So for this analysis, we've chosen a protein called progranulin, it consists of eight distinct domains which make it really amenable to epitope binning. And we have a panel of antibodies to this protein from our wild type bird, that's kind of our bar, and we also have a panel that were generated in transgenic mice, and we've actually compared the antibodies we got from our wild-type bird with mouse, and we see that they're very similar, but chickens do indeed have unique epitopes that they recognize that were not found in the mouse panel.

0:10:21.0 KC: So we've used this protein for years comparing our different transgenes and seeing... Looking at the antibody repertoire of our birds. The meat and potatoes of... This is just a summary of the data from looking at our kappa versus lambda birds. And what's kind of interesting here is, this is basically over 100 antibodies for each different genotype, and we see different biases in coverage. This is a summary of the data from over 200 different antibodies that Yasmina looked at and did epitope binning on. On the left, you have our kappa birds and on the right, antibodies derived from our lambda birds, and there is, it looks like from this cohort, a difference in epitope coverage, it seems like the lambda birds tend to favor the G domain of the progranulin protein, and kappa has more bias towards the CD domain.

0:11:35.9 KC: But I think you have to remember that this is a relatively small cohort, and when we look... I'll show you. When we look at a sequence dendrogram of these antibodies, you can see that within one bin there is a lot of relatedness of these clones, they're probably all derived from a common ancestor and branched out. So, I don't think this is really representing true bias in the genotypes, but rather representative of a clonal type that really took off. And if we looked at hundreds more clones, we probably wouldn't see this. But this is a sequence dendrogram, looking at the sequence relatedness in our kappa with pre-rearranged heavy chain, and you can see what we would expect that clones of similar sequence, as I said, are binning to the same domain. But we also see that there are clones that are really not related, for instance, this P specific clone down here, is really not very related at all to this one up here, so you see different sequences converging on the same epitope bin.

0:12:52.9 KC: Which I think really speaks to the antibody repertoire in our birds. Another thing that can be done on the LSA instrument relatively easily is looking at species cross-reactivity in the same experiment, and in the middle panel here, you can see mouse cross-reactivity for each of these clones, and again, that's kind of clustering with the same related sequences. I should mention that none of these birds were actually immunized with mouse progranulin. The other good thing about chicken is that we can really easily get species cross-reactivity. And the final column on this slides is actually the affinity for each of these clones. And you can see that we have picomolar clones in this small cohort. For comparison, this is one of our... This is the cohort of lambda birds with a pre-rearranged VH3-23 and again, you see the same sequence in binning groupings with mouse cross-reactivity and then affinity on the far right. And again, as I mentioned with these lambda birds, we had complete epitope coverage of the progranulin protein.

  • The Lambda with a Re-Arranging VH3-23

0:14:14.0 KC: This is the lambda with a re-arranging VH3-23. Again, we see really nice epitope coverage from actually just two birds. This data came from... And over half of the birds were mouse cross-reactive in this cohort. So, we're really happy with our lambda birds and the results that we've been getting from them. As I mentioned, Carterra also helps us measure affinity data for our birds, and again, their technology is really great because we can do hundreds of clones all at once in the same experiment, and we can do different comparisons, and it's all done at the same time, which is very nice. So this is an iso-affinity plot of cohort of the kappa and lambda antibodies and their affinities, and you can see there's really no difference between them. We've looked at this several times, they have a range of affinities, some in the picomolar range, and they're both... I don't think there's really a significant difference in the affinities that you get from these different birds. This is just an example of the kinetic diversity that we see, different on-rates that you can actually look at the sensograms from the Carterra data and get a really... A better view of the data than just a raw number, you can really look at the kinetics on these plots.

0:16:01.6 KC: The last bit of data that I wanna show you is looking at the sequence diversity, taking a closer look at that in our birds. These are sequence diversity plots, so on the bottom, you see the germline sequence of our lambda light chain, and on the bottom panel, it's the heavy chain. And each of these colors represents a different amino acid and their frequency in the cohort of antibodies that we looked at. And I think what's really striking, and we see this in all of our transgenes, is that you can really easily pick out the CDRs in these birds, and that is partially due to our design, the pseudogenes have minimal changes in their framework regions, so that we've done to help minimize the troubleshooting that you would have to go through in the manufacturing process, and it's also known that chickens do really have a minimal framework diversity. When you look at NGS data, they don't tend to put a lot of diversity naturally in their repertoire in the framework regions. So, really nicely we can see great sequence diversity in the CDRs but not in the framework regions of our birds.

  • Prearranged Versus Rearranging Heavy Chain

0:17:25.9 KC: So, I keep talking about the prearranged versus rearranging heavy chain, and what is the significance of that? Why did we do that? We did it because we wanted to see different CDRH3 lengths in our birds, and so we've done a comparison here on the left of SynVH-C, these are just unselected from the repertoire, just deep sequencing from SynVH-S... I'm sorry, SynVH-C and SynVH-SD and on the bottom panels are unique antibodies from SynVH-C and SynVH-SD. And you can see that there is a wider range of sequence length, CDRH3 length in the rearranging bird versus the SynVH-SD bird for both the unselected and the unique antibodies in the repertoire. So, that was our intention, was to see more sequence length, and that's what we see when we look at thousands of sequences and hundreds of sequences from antigen-specific antibodies, and that's actually data that's published in Frontiers and goes into a lot more detail also. So in conclusion, we've used the progranulin protein as a model antigen to look at all of our different genotypes of the OmniChicken and to assess their antibody repertoire, and we've seen that using this, that we have really good epitope coverage of this protein from all of our different transgenes. It's very comparable to our wild type birds.

0:19:09.0 KC: We've seen that these genotypes demonstrate bias towards different epitopes in the program and model antigen, but I really don't think that that is a real bias, and I think if we looked at more antibodies and more birds, we would see that difference disappear. Most importantly for this seminar, I hope that I've conveyed that the LSA instrument really allows us to look at hundreds of antibodies together in a single experiment and get tons and tons of data from that experiment. And really the beauty of it is that we haven't purified any of these antibodies, it's the crude supernatant from a 2 ml transfection that we've sent, it's been freeze thawed, the results are really quite consistent when we look over time. Actually, Yasmina just ran some kinetics on some supernatant that she had from us that were two years old, and the numbers were really spot on, there was no difference. I couldn't actually believe it 'cause they were up two almost three years old, and the kinetics came out exactly the same.

0:20:18.6 KC: So, I think that really speaks to the technology of the instrument and what you can do with it. It's really quite powerful to have to be able to look at all of these different clones together and over time. So, I just want to thank again, Yasmina and Dan at Carterra who have helped us generate all of this data. At Ligand, Phil Leighton, he's our director of molecular biology, and he's the one who designed all of these transgenes. Kimberly Berg, who is sadly leaving us soon for graduate school, but she helped generate most of these antibodies that are in this presentation and Ellen and Darlene who are responsible for the primary cell culture that enables us to make these transgenes. So, again, thank you very much for your time and thank you to Yasmina and the team at Carterra for inviting me here this morning.

0:21:20.9 KD: Thank you so much, Kathryn, a great way to start the webinar. And before we move on to our second presentation, I want to remind the audience, you can submit your questions for our Q and A session at the end of the presentations. Okay, we now welcome Yasmina Abdiche from Carterra. Yasmina over to you.

  •  How You Can Use Carterra's LSA Platform to Accelerate your Antibody Discovery?

0:21:43.6 Yasmina Noubia Abdiche: Hi everyone, thanks for joining us today. My name is Yasmina Abdiche, and I'm delighted to talk to you today about how you can use Carterra's LSA platform to accelerate your antibody discovery. So typically, antibody production is pretty highly commoditized in the drug discovery industry. So there's no dearth of antibodies. However, analytical tools to characterize them have, until now, really fallen orders of magnitude behind in that report. What this means is that we're only able to sample really small percentage of all the antibodies that are produced. This doesn't give you a way of really appreciating how much diversity you may have in your panel. So, as analytical tools improve and expand in their capability, we're able now to look more deeply, to screen deeply, to look earlier into the panel and really find those near perfect clones that would require minimal engineering and then accelerate them forward in the drug discovery pipeline.

0:23:06.9 YA: As I mentioned, the concept of doing SPR of high-throughput is rather disruptive. And this enables you to really have this paradigm shift. So, no longer do you have to use SPR only as a secondary tool in characterization, you could use it in screening, and this increases the chances that you're gonna find a really good clone. And this can only accelerate the triage. Our technology is really focused on the application of biophysics to antibodies. And in the antibody space, what you really care about in terms of binding properties are kinetics and affinity and epitope characterization. So, we have a core-driven approach in our software to really facilitate using the instrument. So, you click on an icon, describing the assay format that you want to do and here are a few examples. So on the left, you can see this would be a setup for a kinetic or affinity determination in the middle this would be our epitope binning or competitive binding assay. And we also have really nice software that enables you to analyze these really large data sets quickly and get to an answer that is readily interpretable.

0:24:47.0 YA: Our technology relies on two core principles. We have two microfluidic modules. One is a single flow cell, and what happens here is that the flow cell docks over the chip surface and your analyte is passed over the entire chip, and we call this process knolling. The other mode of action is in the printhead, so we have a 96 channel printhead that docks onto the surface, and it forms the flow channels, perpendicular to the chip surface. And this can nest four times to create a 384 ligand array. And you can imagine there's a choreography between these two microfluidic modules, that enables quite a versatile work flow. Anything that we want to print onto discrete spots, and to dock the printhead, and anything that you would want to expose the entire surface to, you would use the single flow sample. So, I'm just now gonna show a little video of how the 96 channel printhead works. So, we have this 96 meeting manifold, it will collect sample from either a 96 well plate, or a 384 well plate. And it will cycle the sample back and forth across the surface of the chip. It will go and collect more sample and print up to four times to create this nested array of 384 unique spots.

0:26:29.8 YA: Then having been through this array, you could then go to the single close out mode and close out the front over the entire array, and you would pause your sample over all of those spots that you just created. This is a very efficient way of using your analyte. This is great economy of sample, so you have a fixed sample plug of about 250 micrometers, and that sample addresses the entire array. So this would be a one-on-many assay format. This is a close-up of what the printhead looks like, so each channel has its own inlet-outlet, and this stamps onto the gold surface to form a tight interface and these flow cells are perpendicular to the chip. This is a footprint of what the array would look like if you had just addressed one print. So, at the top, in pink, are the discrete spots, and this is where you would immobilize your 96 ligands. So, you could use a standard capture method or mean coupling if you wanted to conveniently immobilize them and then by sequential docks, of the printhead, you could create this nested array of up to 384 spots.

0:28:02.0 YA: In between the spots are areas of the chip that don't get contacted by the printhead, these blue rectangles, these would be your local reference spots. So, when you're in the single-channel mode, your analyte can flow back and forth across the whole surface, and it would be reading SPR the entire time. So, whether the printhead has stopped or whether the single flow service stopped, you're always reading the SPR signal for each of these 384 spots and the reference surfaces. So, when you're in the single channel mode, you can now collect data on both the reaction spots and the reference spots, and then get your reference-attracted data. So, if you had set up a kinetic experiment, for example, you have loaned out your capture antibody using standard imine coupling, let's say it was an anti-human IgG Fc molecule, and then you had printed out a 384 array of antibodies and then go back to the single flow cell and titrate your analyte, this is what your kinetics could look like.

  •  384 Tiles of Kinetic Data

0:29:21.4 YA: So, here I've shown 384 tiles of kinetic data, where each tile represents a unique spot in the array. And what's really nice about this particular picture is that not only is there a ton of data there, but it's a really good illustration of how our software works. So, I like to think of it as showing you the good, the bad and the ugly. So, we have automatic QCs that we've built into our software, to enable you to quickly scout out the good binding Sensograms versus spots that maybe have sub-optimal responses, either because the capacity is too low or that the binding data is heterogeneous and it would quickly flag those spots and create the data for you.

0:30:20.1 YA: So, this is an enlarged view of the data that I just showed you and in this example, there's the same antibody that has been spotted on to 12 different addresses in the array, some of them are adjacent and some of them are non-adjacent. And in this example, we have printed the ligand at three different densities to create high, medium and low capacity surfaces, and the purpose of this slide really is to show you the data quality in terms of reproducibility and how well it conforms to a simple binding model. So here in red, there's a global fit of the data to a Langmuir model, and in blue to green are the increasing analyte concentrations. And so, on a per-spot basis, we apply a simple global fit and because we have more than one spot location per ligand, in this particular example, we can not only just report the kinetic rate constants on a per-spot basis, but we can also report some statistics. So, this gives you more confidence in the reported kinetic parameters.

0:31:45.6 YA: This is another example showing you that in a capture format, we can describe both high affinity interactions and low affinity interactions using the same analyte concentration series. This is a really nice demonstration of sample economy. So by titrating, the analyte from low to high concentration in this particular example, we went from 0.5 nanomolar all the way to one micromolar. We were able to broaden the affinity range that you can access in a single experiment. So, in the top second panels are replicate spots of the high affinity ligand, so we can get to affinity values that are low peak molar range, and at the bottom of this slide are eight spots of low affinity or weak affinity ligand, and that is in the mid nanomolar range. This gives a really broad dynamic range of affinities that you can characterize within a single array.

0:32:57.8 YA: When you have a lot of data, it's important to have analytical tools that can help you to organize that data in a manageable way. So, the data that I just showed you is automatically plotted into either affinity plots, so this would be automatically populated. And here, what's really nice in our software is that you can create ligand sets. So for example, let's say you've had two different libraries, you could assign some of the spots to be library one, some of the spots to be library two, and then in this kind of figure what shapes the distribution of the kinetic rate for instance, you could color those libraries differently, and this will enable you to quickly see whether one library had an affinity bias than the other. In our software, we have default cut-ups that we apply to the Koff rate so that it doesn't give you a ridiculous value. So, in this particular example, we've been rather conservative and made the cut-off of one eight to minus five in the Koff rate.

0:34:14.1 YA: And so in this kind of plot, that's why I've shown here with the green arrow, you get a clustering of antibodies that meet that limit, these are interactions that are too slow to be measured accurately during this particular dissociation phase. When you have less than 384 in each spots, why not print multiple replicants of each clone and then increase the end, like I mentioned before. In this particular example, we have 34 unique anti-bodies, and capture them on an average of about 10 spots each, and then we can generate Mean and Standard Deviations of each of these interactions, and this is automatically generated in our software. So, this would be the koff rate and this would be the data for the affinity. So, when there's a new technology, it's really important to benchmark against existing tools so that people can understand how our technology fares in terms of data quality and reliability.

  • Example of a Collaboration That We Were Doing With Adimab

0:35:34.7 YA: Here's an example of a collaboration that we were doing with Adimab, where we compared the same set of clones in the same set up, so the data that I just showed you, which was kinetics determined on a capture surface. We repeated the experiment on a BFO 8K and got remarkably consistent results across the two platforms. So, when using a flat chip type on the BFO and using a flat chip type on the LSA, we were able to generate a correlation, that would be as good as doing the experiment two independent times on either platforms. As you can see from the affinity range here, we were spanning four orders of magnitude and still getting a really tight correlation, and these data were generated 50 times faster than on a BFO, so for every eight interactions that are processed on the BFO we're able to generate 384 interactions.

0:36:41.7 YA: And because of that one-on-many assay configuration that I related to earlier on the single closed cell side, we consumed very little sample, so less than 1% of the sample that you would need on a BFO. Bear in mind that, we were able to generate data of 384 different spots pretty much in an afternoon's work, and it would take much longer on a BFO to get the same capacity. And in addition, we have this really powerful analytical tools and the software that enable batch fitting and really fast access to the results. Here's an example from some caricomsmic extracts, where these are antibody fragments, SCFPs. This example really highlights the excellent data quality that you can get even from a crude samples. So using the print head this really acts as an inline purification, so in this particular example, these SCFPs had been engineered with a V5 tag, and so using an anti-V5 capture nor we were able to use the pinhead to capture each of these clones onto the anti-V5 and then in a single flow cell titrate the antigen. As you can see, excellent description of their kinetics to a simple model.

0:38:22.3 YA: In this particular library, we were able to see interactions that were single-digit nanomolar all the way to single digit micromolar. I'm gonna switch gears now, and I was telling you about kinetics and affinities. Now, I'm going to talk about epitope and why it's all about the epitope. So, with therapeutic antibody discovery, where an antibody binds is going to dictate it's mechanism of action. And not only is it important in dictating how the molecule is gonna work, it's pretty much the only aspect of an antibody that you can't engineer on purpose. So, I'd like to think of an antibody as being born to be specific to a particular epitope. So, it's an innate property, you can't shift it rationally, it can't be predicted in silico, so you have to rely on empirics selection. So, not only is it important to the biology of your target of how the interaction is gonna proceed, but you can generate some intellectual property around it as well. So, IP around epitope is highly regarded in that discovery.

  • Try and Find a Functionally Relevant Mechanism Where the Epitope is a Little Bit Different

0:39:45.4 YA: And so, you always want to try and find a functionally relevant mechanism where the epitope is a little bit different. So, high throughput SPR can really help you to cluster the antibodies into epitope families, to distill the panel into a subset, that collectively represents all of the epitopes. So, I mentioned earlier, epitope binning, this is a competitive binding assay, and we do it in a pair-wise combinatorial way and just ask the question, can these two antibodies bind the target at the same time, and form a sandwich complex or does the binding of one antibody block the binding of the next antibody? So, by looking at every pair-wise permutation if you have a large kind of antibody, this very quickly becomes a large asset, it's a geometric scaling.

0:40:51.5 YA: So, say if you have 384 antibodies to pair them against one another, you would be looking at about 147,000 pair-wise interactions. So, this really is beyond the capacity of any other analytical tool, is really a unique way in which the LSA can really give you a lot of information that you just couldn't get another way. So, I'd like to spend a few minutes really looking at these two diagrams at the bottom. So, in A, on the left, is an example from the literature that I pulled from 2016, where these authors used a 13 by 13 epitope in an experiment using the Octet biosensor, which is a very commonly used technique for these kind of experiments. I chose this example because it's really representative of the size of panel that people are using when they're using other techniques. And in the same year in 2016, we published this paper in maps using the 384 array capability.

0:42:02.8 YA: This was the first demonstration of how you could do these really big binning experiments. It's pretty much contrast, these two images on the left, it's kind of like black and white TV, and on the right, it's like your high definition color pixelation. And the point here is that when you have more pixels in the picture, you build up much more granular, more detailed view of what's going on. This enables you to really have a high resolution, high definition in your antibody panel, and find unique epitopes that would otherwise be overlooked. So, in our software, we have some really nice tools, this example shows a community blocs. So, there's two big blocs there, the purple bloc and the cyan blue community. And this really is a very course demarcation of how these epitopes are clustered into two main communities, and we're gonna come to this in a moment. So, building from what I just said, I wanted to explain a little bit more in detail about this large binning study that we published down in 2016.

  •  Purpose of Using an Antigen

0:43:25.0 YA: So, for the purpose of using an antigen that had been very well described in the literature, we chose hen egg-white lysozyme. When we looked in the literature, an aggregate of 20 years in published work, it had the most number of crystal structures, the complexes of antigen antibody. And there were about 79 when we looked in the literature, and of those 79 structures, they represented about 20 unique antibody sequences. Of those 20 unique antibody sequences, we picked five antibodies, that collectively we distributed over the greatest amount of surface of this antigen, and this is shown on the left, controls one through five. And this is... They were kind of scattered across what we will call the front face of the hen egg-white lysozyme. But if you turn that structure around, you can notice that there's this bald patch at the back, and we can speculate that possibly, that's because the antibodies that have being described in the literature were mainly from immunization. So, it's possible that the front face of how is more immunogenic than the rear face. So, it offers this naval binding spot that hadn't been accessed before in the literature. So, what we did is we took the Adimab library that had been raised against...

0:45:00.0 YA: Hand against hen egg-white lysozyme, and we did a big binning experiment and we asked the question, which of those control antibodies were being blocked or not by each of the clones in the Adimab library. And we assigned a bin name or blocking profile to each of the clones. So for example, the clone that was not blocked by any of those controls that I just showed you, would be assigned for blocking profile zero, and it's possible that, the reason that it's not blocked is that it's binding to that rear face of hen egg-white lysozyme. And another example would be blocking profile one, that would be an antibody that is only blocked by control number one and not blocked by any of the other controls. So, we built up a set of distributions based on these different blocking profiles, and we looked at the literature and there's 20 antibodies that I mentioned, and in an Adimab library about 214 antibodies only assigned the blocking profiles, and then the next slide will show you how these look in the venn diagram.

0:46:11.4 YA: So, on the left here is the distribution of antibody binning profiles that we had identified in the literature, and on the right, the antibody binning profiles that we identified in the Adimab library by applying empiric epitope binning assays. And basically the take-home message here is that by screening very deeply in doing this high throughput binning analysis, we were able not only to find every epitope that had been described in 20 years of literature on this particular target, but also discover new epitopes that hadn't been described previously. So, this really is a great example showing you that if you look deep enough in your library, you're highly likely to find rare bins that maybe could offer you some IP advantage or some unique opportunity. So, high throughput binning as enabled by the LSA, is really transforming the binding paradigm, 'cause you can't really do these assays on the other available technologies.

0:47:27.4 YA: That's really because doing these assays just take too much material, it's very high on sample consumption, when you have a lot of sample, you need to plate it out, that's very onerous and you would need pretty much a robot to do the liquid handling. You typically wouldn't have this much antibody available on so many clones earlier in a drug discovery program, and so these assays until now really have been overlooked. But the LSA is really helping to transform the binning paradigm and enable researchers to understand the epitopes diversity earlier. In the software, we have these three different views that pop up; on the left is the sensorgram view, and this is always connected to the heatmap, and then these network plots that are derived from the heatmap are shown alongside. So, if you click anywhere on any of these three panels, the relevant data in the other panel shows up, so you're always connected to the primary data.

  • Interrogate Any Particular Binning Assignment

0:48:42.0 YA: This enables you to interrogate any particular binning assignment that you may be querying, and then you can see the data and make the judgment by yourself, whether you believe it or not. Along with the visual nature of our epitope tool software, we have alternate ways of looking at the data. So, in the middle here is a dendrogram, and this view is commonly used by people who analyze sequences to see how related different sequences are. So, in the context of binning, this is telling you how similar they are in their binning or blocking profiles. So, on the left is the most granular way of interpreting the dendrogram where the networks are taking a cut height at the very tip of the dendrogram, so this allows for no differences between the blocking profiles of each clone.

0:49:43.0 YA: On the right, we've allowed a little bit more forgiveness, if you will, and said that we are allowing a few differences in blocking profiles, so this gives a closer view and generates communities. So, I hope I've shown you today that the LSA truly is a disruptive technology when it comes to SPR. So SPR is known for its reliable binding kinetics and real-time profile, so you're able now to do that in high throughput and use SPR early in your screening. This allows you to do screening characterization as the same step, and the rate constant that we're getting are going really with BFO, but we're generating them much faster and with less than one percent of the sample consumption. Thank you so much for your attention today.

0:50:46.5 KD: Thank you very much Yasmina, another fabulous presentation. So before we start the Q and A session, I want to let you know this is your final chance to submit your questions for our speakers, so please get on with it and submit your questions now. Alright, we have some good questions in the queue, so we will try to get through as many of these as we can in the next short while. Let's get everything situated and we'll transition into the Q and A session.


0:51:28.3 KD: Okay, everyone. We've got time for a few questions. Kevin Davis back with you. So I think the first one, Kathryn, will go to you. It's another Kathryn who asks, "Are you able to achieve picomolar affinities for carbohydrate antigens using the chicken? And what type of sequence diversity have you been able to achieve using carbohydrate antigens?"

0:51:53.6 KC: Oh, that's a great question, Kathryn, and a very, very great name. We actually haven't done a ton of work with carbohydrate antigens. We have done some. I think the challenge there is being able to get a good quality preps, and the project that we were working on, we were just really never able to achieve that. We did get a titer in the birds that we immunized, but we really... Creating screening agents was a challenge for the project, so that wasn't actually something that we pursued very far. I think with the right partner and the right reagents, we probably could do a lot more work. But typically with our programs, we are able to achieve picomolar affinities for, I would say, the majority, the vast majority, of targets that we look at. And we've published quite a bit on sequence diversity from our birds. We have a matte paper that's about two years old, and also a Frontiers paper, taking a closer look at the sequence diversity from a couple of different genotypes. And, as I mentioned in the talk, most of that sequence diversity is really concentrated in the CDRs and not the frameworks.

0:53:09.4 KD: Okay, thank you. I think our next question is from the same person, but Yasmina, I'll put this one to you. First off, "It appears that SPR technology is optimized for soluble antigens or analytes; however, many targets are not amenable to being used as an analyte. So how do you deal with those types of antigens, membrane proteins, aggregated membrane proteins, or non-protein antigens? Can you invert the assay and get higher throughput compared to existing screening platforms?"

0:53:43.2 YA: Thank you for your question, Kathryn. So yeah, this is a great point. So our technology is SPR-based, so it's limited by the caveats for SPR, which as Katie alluded to, is very reagent dependent as most assays are. So if you have a poor quality reagent, you're absolutely right, it's not going to function well as an analyte. In cases such as peptides where you have poor solubility or thay're oligomeric, you can immobilize them if you have, for example, a biotin on one end. Typically for SPR assays, you want to put the multimeric partner on the surface. And so in our technology, because you can look at 384 different ligands in parallel, you could use that for scouting your antigen preps and therefore achieve a higher throughput than other screening platforms. But in general, poor quality reagents are not gonna be amenable to a proper SPR kinetic analysis, and there are some methods out there for improving reagent quality. So companies, such as Integral Molecular, do a great job of preparing membrane proteins and just reagent preparation in general. It's very worth while effort, if you want to get good data. Thank you.

0:55:13.4 KD: Thank you. Thanks, Yasmina. We've got time for just a couple more questions before the top of the hour. I'm gonna stay with you, Yasmina, for this question, which comes from Gwen Wei, who says, "I've noticed that on the current single concentration kinetics cleaning mode, there's no such an option to add warm-up cycles before the real sample injection. I hear that this feature will be included in the new version of Navigator. Can you give us a rough idea when the new app will be released?"

0:55:45.5 YA: Well, thank you for your question. So there are various options available, not just in the core apps of kinetics. We have the quick startup, but also you can choose to have a regeneration or not. So depending upon the assay format that you want, you're not limited to using injections that require a regeneration step. Thank you.

0:56:12.3 KD: Thanks, Yasmina. Let's go back to Katie, as we reach the close of our webinar. Katie, you talk mostly about the chicken. What other kinds of birds are you planning to make? 

0:56:24.1 KC: Oh, that's a great question. We actually just introduced our omni quick chicken, which is a fixed light chain bird. So this is a bird that's designed to be used for bispecific antibody discovery. And we've literally just hatched them recently.

0:56:45.5 KD: Oh, great.

0:56:46.5 KC: And we've been looking at... Yeah, we've been looking at how they use that fixed light chain, which I'm pleased to say, we see minimal diversity in that fixed light chain and also a lot of clones that we've pulled out that carry the germ line light chain. So we're really excited to get started on some projects using that bird for a bispecific discovery.

0:57:07.1 KD: Okay. And I think for our final question, I'll put back to Yasmina. Yasmina, maybe you could just clarify what is the auto sampler capacity of the LSA, and how many ligands and analytes does it support? 

0:57:22.7 YA: Thank you for that question. Yes, so currently, the LSA has two plate positions that serve the single needle, the single slow cell. Each one of those plate positions allows for us to create four well plates. So typically, in an assay, if you wanted to have a regeneration, you would have a sample block in one of those two bay locations and up to 384 different analytes. And then on the ligand side, on the print head side, we have three plate positions and each of those can hold a 384 well plate. So an unattended run would enable a reload of 1152 different ligands. Thank you.

0:58:08.5 KD: Thank you. Unfortunately, we're out of time, but a very informative webinar. I want to thank both of our speakers. And for those of you who asked a question that we didn't have time to get to, please reach out either to our two speakers or to your Carterra contact or representative. I'm sure they'll be happy to take and address your questions off line. Alright, our webinar today will be archived for another six months on the GEN website. So please, you can watch it on demand or forward the link to colleagues who may not have been able to catch it today. I'd like to thank our two special speakers, Kathryn Ching from Ligand Pharmaceuticals, and Yasmina Abdiche from Carterra for doing such a great job; Carterra, for sponsoring this webinar, but most of all I'd like to thank you, our audience, for making it this far and staying for the full hour. Thank you so much. Hopefully, we'll see you all again for another GEN webinar in the very near future, but for now, for the entire GEN team, this is Kevin Davis, goodbye for now.