Bispecific antibodies play an increasingly important role in the therapeutic antibody space today.
A major challenge in the field is the assembly of the correct heterodimer once two antibodies have been identified. Ligand Pharmaceuticals has addressed this problem by engineering a chicken that expresses a common light chain, thereby conferring epitope specificity entirely on the expressed heavy chain. As a host animal for immunization, the chicken offers many advantages including the ability to recognize highly conserved targets that are poorly immunogenic in mammals.
Presented by Kathryn H. Ching, PhD, Research Investigator, Ligand Pharmaceuticals
0:00:00.4 John McKinley: Welcome, everybody, and thank you for attending today's webinar. During the webinar, please send your questions via private chat to me, John McKinley, the host, and they will be answered at the end of the webinar. Carterra is helping many of our customers screen panels of antibodies in the race to find cocktails of therapeutic antibodies to combat COVID-19. We also have a new nickel biosensor chip for the LSA. Today, we're privileged to have Kathryn Ching from Ligand Pharmaceuticals present on the success of their bispecific antibody discovery platform and workflow. Kathryn works on the design and characterization of the various transgenic lines of the OmniChicken, an animal carrying fully human V genes at its immunoglobulin loci. Before joining Ligand in 2013, Dr. Ching completed post-doctoral fellowships at the National Institute of Dental and Craniofacial Research and the US Department of Agriculture where her work focused on the use of antibodies to diagnose and monitor disease. Dr. Ching received her PhD in biochemistry and molecular biology at Georgetown University Medical Center where she studied role of small G proteins in immune cell activation. Kathryn, I'll hand it over to you now.
0:01:25.0 Kathryn Ching: Thanks, John. Thank you, especially to the team at Carterra, for inviting me to share some data with you this morning. We've worked with Carterra ever since I started at Ligand, really in close partnership. We've published a lot of papers with their scientific team and I'm really grateful for all the help and guidance they've given us in characterizing our antibodies. And I'd like to really touch on our new bispecific antibody discovery platform that we call the OmniClic chicken.
0:02:04.0 KC: So as John said, just as a reminder, if you have any questions at any point during the webinar, please send them via the private chat window to John and we'll try and answer as many of those as we can. Noah Ditto, who is their technical guru, is also on the line, so if you have any technical questions about the instrument, he will also be there at the end to answer some of those. Just about the OmniAb platform that is part of Ligand, it's actually five platforms in one. When you license the platform with us, you get access to five different transgenic animals, three different species, the OmniRat, the OmniChicken and the OmniMouse. The OmniRat, OmniChicken and OmniMouse are animals that carry fully human V genes, as John said, and produce fully human sequence antibodies for classical heavy by light chain antibody discovery. The OmniClic and OmniClic animals and the chickens are specifically engineered for bispecific antibody discovery and they also are carrying fully human V genes at the IG locus. So with an OmniAb license, you have access to all five of these animals for your discovery, antibody discovery needs.
0:03:26.7 KC: I just wanted to give you a brief overview of the chicken. A lot of people aren't familiar with why we use chickens for antibody discovery, and there are a lot of different challenges, as you'll see in the talk, with using a chicken, but chickens are more phylogenetically distant from humans than from mice, and so a lot of times, people come to use the chicken because you're able to raise an immune response to really highly conserved mammalian proteins that you often can't see an immune response at all in a mouse platform. So a lot of the people that we work with come to us for that reason to work on cell surface receptors and other really highly conserved mammalian proteins. So that's one reason why people go to the chicken, but chickens are a little bit different. They do do V [D] J recombination but they don't actually generate a lot of diversity from that recombination because they only have a single V gene at the light chain and heavy chain locus. They rearrange those genes and then they generate diversity through a process called gene conversion. And so what that involves is an upstream array of pseudogenes that will donate stretches of sequence to the rearranged V and in that way, they're able to generate diversity.
0:04:47.0 KC: So these pseudogenes are promoter-less and they don't contain recombination signal sequences, so they themselves cannot rearrange. It really is a process of a recombination that is quite different. So just in cartoon form, you have the rearranged V, the promoter, and then this array of pseudogenes with different sequences in their CDR. And in the first gene conversion event, you'll get some transfer sequence and some diversity, and it's an iterative process, so it'll happen again and again as the B cells undergo development. So what's neat about this is that because there's only a single V gene in these chickens, we were able to pick a single framework, and that really simplifies the downstream PCR process 'cause you're not dealing with primer cocktails of 10 or 15 different primers, you just have a single framework that you need to amplify. So it makes it a lot more efficient. So the question is, what V gene do you choose? And ultimately, we did a lot of in vitro expression studies, also looking at NGS data to see what V regions are common in the normal human population. And so for our heavy chain, we settled on VH3-23, which is really highly expressed in the normal population, and we have two versions of the heavy chain, SynVH-C, which is a pre-rearranged version, and then what we call SynVH-SD, which is VH3-23 but it also…
0:06:21.5 KC: It's a rearranging transgene containing all the human Vs and a single J. And those heavy chains are then paired with either VK3-15 or a VL1-44. And so in essence, you have four different combinations of transgenes that are available on the OmniChicken platform. The other thing that's neat about chickens, because they do this gene conversion process, they don't actually generate a ton of diversity in the framework regions, so most of the diversity, as you'll see later in the talk, is concentrated in the CDRs, and that is really relevant when you're trying to really optimize the production of your antibody. So for every transgene that we have introduced into the chicken, we evaluated the bird using what we call our model antigen and we've chosen a protein, the progranulin protein, which is a multisubunit protein, and we've done this because it's really amenable to epitope binning. It's also readily available commercially. We buy it in buckets from Sino and R&D Systems, and it's highly soluble, so it's really easy to work with. So we've immunized all of our transgenics and a cohort of wild-type birds with this protein and we've used the LSA from Carterra to compare the antibody repertoires of birds immunized with this protein and really setting the wild-type chicken as our benchmark early on.
0:07:58.2 KC: And we published on that in 2016, finding that our first generation of OmniChickens really had a very similar immune response to wild-type birds, and I'll show you some of that data in this talk. So just a quick look at what kind of titers we get from these birds, on the left, you have a kappa expressing bird and on the right, a lambda expressing bird. And you see right at draw one, they have a really robust titer to the progranulin protein that continues to be maintained and even increased throughout the immunization. These were some of the first birds that we did. So we don't usually do seven or eight booths in our programs but we just wanted to see how long the bird could maintain those titers, so this is some very, very early data. As I mentioned, chickens are a little bit difficult to work with for a number of reasons but one of those reasons is they don't have... There's no B cell fusion partner for chickens, so there really was no easy way to maintain heavy and light chain pairings in chickens until a team of scientists at our company developed what we now call the GEM technology, and this is basically a micro-droplet technology that allows us to isolate single B cells. So what we do is we harvest the spleens from immunized animals. We encapsulate individual B cells in these micro-droplets, so this little green guy right here, and encapsulated in that micro-droplet are also beads, polystyrene beads, coated with antigen.
0:09:43.5 KC: We can also do cells expressing the antigen at the cell surface. That's really quite a flexible system. So we incubate these single B cells with antigen-coated beads and a fluorescent secondary antibody for a few hours. So the cells will secrete antibody and if the antibody is specific to that antigen, it'll bind and concentrate that fluorescent signal at the bead, and we can actually see this under the microscope. So this is an example of a cell with an antibody recognizing that antigen lit up in red with the cell there in yellowish green. And all around it, you can see there are these GEMs with cells in them and beads but they're not lighting up. So it's really a neat way to screen upfront for either functionality or just antigen specificity, which any of you who have done hypergamma know that you have to wait at least a week to get that initial screen. And I've been asked by my colleagues at work to point out that this picture was taken long before we had coronavirus amongst us, so we do not sit this close to each other now in the lab anymore. But these are your two scientists at Ligand screening GEMs in a... One day we can screen 10 to 20 million B cells from the spleen for antigen specificity, so it's pretty high throughput.
0:11:19.5 KC: So you're all here today to hear how we have used the LSA instrument in our studies and analysis of panels of antibodies that we get with using... We use actually the generation before the LSA when they were developing the technology and it's really brought another level to the information that we're able to get from our chickens. So if you don't know, LSA stands for Lodestar Array and the idea there is that, just like you have hundreds and hundreds of antibodies that you wanna look at, there are millions and millions of stars in the sky and you're really looking for a tool that will help you find that one antibody, that one star, that you really want as a therapeutic. So it's kind of a clever name, I think. So there are a number of different things that you can do on the LSA and just briefly, one of the main things that we do are kinetics, on and off-rates, to calculate affinity, and that would involve immobilizing your antibody of interest to learn their proprietary chips and flowing the antigen over, looking at on and off-rates and then doing curve fitting. So just an example of some of the sensograms that are generated.
0:12:38.0 KC: The neat thing about the LSA... And I'll probably say this a million times during this talk, but because it has such a high-capacity, we're often able to run our antibodies and duplicate and triplicate at different concentrations and so we really have confidence in the accuracy of those results in kinetics, which anyone who has done kinetics, it's pretty challenging to do. The other main thing that we do with the Carterra is epitope binning, and you can do this in either those two ways. One classical way is immobilizing your panel of antibodies, adding the antigen, letting it bind and then flowing over that entire panel of antibodies over the chip and looking for whether or not the antibodies block or cross-block one another. You can also do it in a premix format where you mix the antibody that's gonna be in the flow with the antigen and then add that to the chip. So we've done that both ways with some success. This is some of the data that can be generated with their software, and I'll show you more of that later. Finally, you can also do peptide mapping, epitope mapping with the LSA. That's not something that we've done but it's definitely possible.
0:14:03.2 KC: So what exactly do we do at Ligand? Well, the scientists at Carterra have actually developed this clever assay to aid in the epitope binning of the progranulin molecule and what they did was... You can see here we have full-length human progranulin and we also have mouse progranulin, which is similar but not identical. They generated a bunch of different chimeras. So chimera 1, you'll have the P and G domains are the human domains of a molecule and the rest is mouse. Chimera 2, S, B and A are human but the rest is mouse. And so we're able to include these chimeric proteins in a number of what we call standards, so antibodies that we know what bin they are assigned to on the chip in the array to help us tease out what epitope bins each antibody in our panel might belong to. And this is an example of the 384-well array that we use. Okay, so this is an example of the kind of data that you'll get out of the instrument. So on one axis is the antibodies that are immobilized on the chip and on the other axis are the antibodies that are in the flow with the analyte. And the red square represents a blocking relationship between two antibodies, whereas green means both antibodies are bound at the same time. And what's cool about this figure is that you can see really easily that there are different clusters of antibodies, different bins that antibodies will be assigned to. The black squares are basically self-blocking, so that's neat.
0:16:07.5 KC: Luckily, with their software, you do not have to stare at that heat map to try and discern what you're looking at as the software generates these really neat nodal plots. So this is some early data in which, as I mentioned, we wanted to compare the OmniChicken antibody response with a wild-type chicken. And so what the software does to de-convolute the data is make these nodal plots and each of these bigger blobs represents one of those domains on the progranulin protein. And you can see in blue is the wild-type chicken and in red is the OmniChicken and you can see that both chickens hit every domain on the molecule. The lines, I should mention, they're representing a blocking relationship between two antibodies, so much easier to look at than the heat map. We frequently take that data and we can line it up against what we know of the sequence of the antibodies that we get. So this is a phylogenetic tree of antibodies based on their sequence and you can see the different families and the relatedness of each clone matched against epitope binning and mouse cross-reactivity data, and finally, kinetics, which I'll show you in more detail in a second, but it's a really nice way of just summing up the data. You can see, obviously, that sequence is better related, bin into the same epitope bin. We also get some sequences that are not very closely related, like this G3 antibody down here and A9, it looks like, in the P domain are not very related sequence-wise.
0:18:02.6 KC: I do wanna mention one neat thing about the chicken. Because of its phylogenetic distance, it's really easy to get species cross-reactivity. So you'll see that half the antibodies in this panel are mouse cross-reactive, and we actually did not immunize at all with the mouse progranulin protein. We see that, quite frequently, we're able to get pseudo mouse cross-reactive clones with relatively little effort.
0:18:31.0 KC: This is just another way of looking at the sequence, the sequences of these clones. So on the X-axis is the germline amino acid sequence of the light chain on top and heavy chain on the bottom. And as I was saying, chickens, there's very little diversity in the framework regions and you can actually pick out the CDRs quite easily. So each one of these colors represents the frequency of a different amino acid at that position. So in the chicken, antibody diversity really is concentrated... I'm sorry, amino acid diversity is really concentrated in the CDRs. So this was, I think, about 100 different antibodies and this was published in labs when we described the first OmniChicken. As I said, affinity is also something that you can get using the Carterra and again, we run these in duplicate, triplicate at different concentrations. So this is an isoaffinity plot with increasing affinity going towards the northwest and we get many high-affinity clones to the progranulin protein. And again, really, without doing any optimization to the immunization protocol, we can get sudden animal or affinity from our birds. So that sort of sums up a lot of the work that we've done.
0:20:03.5 KC: For the OmniChicken, looking at epitope coverage of the different transgenes that we used for heavy by light antibody discovery. We get sudden animal or affinities for the progranulin protein, but also quite commonly for a lot of our other programs. And the LSA instrument has really helped us understand the repertoire of our antibiotic panels for this model antigen and given us a lot of information that we wouldn't otherwise easily get, especially the epitope binning. I think that's a real asset to be able to run so many different antibodies at once, and you'll see that in another slide later on. If you're interested in any other more technical information, we have a few different papers describing the kappa birds, the lambda birds and an NGS analysis of our heavy chain, looking at differences and similarities between our pre-rearranged and our rearranging heavy chain that contains all the human Ds. So those are open access, if you're interested.
0:21:12.7 KC: So, now I wanna sort of transition over to tell you about our OmniClic bird, which is intended for bispecific antibody discovery, and this is a bird that is fairly recently hatched. [chuckle] We're working on a manuscript right now describing it, but we're really excited to get it out there and share it with partners and see how it performs. Excuse me. So if you're not familiar with bispecific antibodies, one of the great challenges of it is that you have, in the simplest format of an IgG, you want a molecule where one arm, one heavy and light chain has specificity for antigen A and the other arm of the molecule has specificity for antigen B. You have four different proteins that need to come together with the correct pairing in order to get the bispecific antibody you want. So obviously, you're gonna have some problems because you could have homodimers of the heavy chain instead of heterodimers, and you can also swap light chains, so you could have antigen A's light chain with antigen B's heavy chain and vice-versa. So getting the right pairings for bispecific is really a challenge in the field, and we sought to make a chicken that would address that challenge.
0:22:42.3 KC: So our idea was to create a bird where antigen specificity was entirely driven by the heavy chain, and so every antibody that is generated from the bird, we hoped would express what we call our common light chain, so every antibody, every heavy chain would pair with the exact same light chain. So you would not have to worry about light chain pairing in the end point molecule. So how do we do this? If you remember with our kappa bird, our diversifying light chain transgene in the OmniChicken, we have a pre-arranged VK pseudogenes upstream of an array of pseudogenes, based on 3-15 but with all different sequences in their CDRs. And that creates what you saw on those different sequence plots really, a lot of diversity in the CDRs. So the idea with the common light chain bird was to have that same VK3-15 because we see great expression in our birds, but instead of a diverse array of pseudogenes, to have all identical pseudogenes. So when you go through that gene conversion process, gene conversion will occur but in theory, you won't actually be introducing any diversity into the rearranged V because there is no diversity in the pseudogene array. We do however... Chickens do somatic hypermutation, so we do see some mutation that occurs in the periphery because of that but it's very minimal, as you'll see.
0:24:25.7 KC: So, these birds, one of the first sort of metrics that we looked at was whether or not they could produce mature B cells with this fixed light chain, and indeed they do. This is the FACS self-sustaining of wild type versus birds expressing the common light chain. And you can see there really is no difference, B1 is a chicken B cell marker, there is no difference between wild type and the common light chain birds. We see surface Ig expression, heavy chain and light chain, we see normal populations of T cells, and this antibody was an antibody that we raised in rabbits against our germline V gene, and so it should only recognize a transgenic animal, which you can see quite clearly. There's no staining on the wild type bird and quite a bit of staining on the light chain, the OmniClic B cell population. So that was very promising. We then took splenocytes from these birds, unimmunized, and performed NGS sequence analysis on them to see sort of what the population looks like, and you can see on the left this is NGS, an OmniChicken cohort of birds. In blue is the light chain, in red, the heavy chain, and you can see the number of changes compared to germline, it's very similar for heavy and light, an average of 9.6 versus almost 12 for both the heavy and light chain, so quite a few number of changes.
0:26:10.0 KC: When you look at the OmniClic bird, you see a very dramatic shift in the light chain numbers to the left. Here in the OmniClic bird, the average was about 3.5, while on the human heavy chain, we're still quite seeing quite a number of changes compared to germline, and actually there, a significant number of birds were just expressing the germline light chain, so this is pretty exciting for us to see that it seems like our strategy really of creating a non-diversifying transgene was actually working. The birds were able to mount an immune response against our favorite progranulin protein. These are two birds, on the left is one with a pre-rearranged heavy chain, and on the right, one with a rearranged, and you can see that they both mount a really strong immune response, and in fact, we sacrificed them after only three booths and took their spleens. So coming back to the Carterra now, looking at epitope binning, on the right is the panel of antibodies generated from the OmniChicken, and on the left, the panel of antibodies generated from the OmniClic bird, and you can see that we have complete epitope coverage in the OmniClic bird, it looks very similar to the regular OmniChicken.
0:27:39.3 KC: A couple of interesting things I wanna bring up in this slide; one is that the antibodies on the right from the regular OmniChicken were actually expressed originally for the paper that we published describing that bird and then frozen down and kept in the freezers for a few years. And we weren't sure, but we thought we'd give it a go. We saw them and we ran them in this experiment side by side with this new panel of antibodies termed the OmniClic bird, and quite nicely actually, they all binned to the same epitope bin, and the kinetics were very consistent with what we had seen originally. So I think the reproducibility in the instrument is really... It was a nice thing to see even with supernatants that were obviously, I mean, they were frozen, but they were still several years old. So it was really great to see that reproducibility on the instrument. The other thing that I wanted to mention is, I don't know, if you go back and recall the nodal plot that I showed you earlier, all the different sub-domains were just kind of a distinct blob, and what we found as we've generated more antibodies to these birds and we've run more antibodies on the LSA instrument, is that we've been able to tease out within a sub-domain, we've been able to tease out maybe some differences in binding within that domain.
0:29:11.9 KC: So here, with the B sub-domain, suddenly we're seeing different pockets of antibodies that interact differently in comparison to one another, and that is, I think... I mean it's only possible because we're able to run so many antibodies at a time, and we've been able to select antibodies that we can use a sort of standard benchmark, and we're able to tease out a lot more information about sub-domain, what epitopes are being recognized by different antibodies. So again, and I know I keep repeating this [chuckle] but really, the high-throughput nature of the instrument allows you to get this data. It's really nice.
0:30:06.1 KC: This is another... The same plot looking at the number of changes compared to germline, but this time with antigen specific clones instead of just bulk splenocyte. So on the left again, you see a panel of antibodies from our regular OmniChicken with a diversifying light chain, and you can see a quite similar numbers of changes between heavy and light, and then if you go over and look at the OmniClic bird, again you see that shift to the left with the light chain. And really, in this instance, quite a few antibodies within the panel just expressing that germline light chain, so it's nice to see the data from the NGS kind of translate that to the specific antibodies that we pick up in our bird when we do... Actually screening for antigen specificity. Again, we took that panel of antibodies from the OmniClic bird and we did kinetics using the LSA, and another neat thing about the software is that you can color code... Sorry, that's my dog. You can color code the different antibodies based on different bins, so what genotype it came from or whatever you like, in this instance, we've plotted affinity and color coded according to epitope bins, so you can see the different affinities for the different bins, so a different way of looking at the data, but very, very useful.
0:31:49.6 KC: We also looked at affinity, we were able to measure in the same experiment, clones that were carrying a native, what we called a native V-kappa, so a light chain that did indeed diversify most likely through somatic hypermutation versus clones that were carrying the germline, a completely germline. And there is a slight difference in the mean in affinity, but it actually isn't significant, so we were pretty happy with that. You would expect that maybe the affinity wouldn't be as great, but it's nice to see that there's no statistical difference there. We did a little experiment where we took a number of these clones that had some changes from germline in their light chain and we re-cloned them with the germline light chain and re-expressed, and then we used the LSA to measure them side-by-side. So this line has a slope of one. On the X-axis is our clones, the affinity of clones carrying their native light chain and then on the Y-axis, those same heavy chains, but now with the germline light chain, and you can see there really wasn't, except for maybe a couple, a significant change in their affinity, and we did binning again with these. And they were assigned to the same bins, so that's really promising.
0:33:13.6 KC: It shows that the strategy is having a non-diversifying light chain and driving antigen specificity by the heavy chain is really working. This is one of those phylogenetic trees again, grouped by sequence relatedness. In green, these are clones that has the germline light chain, non-diversified and then again, next to their epitope binning and mouse cross-reactivity. Excuse me. Just another way of looking at the data, looking at different sequences and how maybe different amino acids are affecting affinity, or possibly the different epitope bin assignments. This is a plot, again, of amino acid diversity on the light chain on top versus heavy chain on the bottom for our OmniClic birds. And again, we usually don't highlight the CDRs when we make these plots because it's so obvious where they are like on, the heavy chain on the bottom there, if I didn't draw a box around that, you'd probably be able to pick out the CDR, but we felt like it was needed in this case, because the light chain, there really is so little change that's occurring in the bird. Excuse me. That we highlighted the CDR so that you can see them. A lot of these clones, if I go back a slide, some of them are really closely related, so even this amount of change, it probably represents clones from the same family that are really not that far apart, so in reality, I think, the strategy of using an identical array of pseudogenes is really working quite efficiently.
0:35:10.6 KC: So in conclusion, I think I've shown you a lot of data from both the OmniChicken and the OmniClic to show that they really have, against our model antigen, a really good epitope coverage on our model antigen. The OmniClic, I think is going to be a really valuable tool in the community for bispecific antibody discovery because it really does focus antigen recognition on the heavy chain, and not the light chain. We're really excited to see that bird go out into the world and see what it can do. That is all that I have. I just want to thank the team at Ligand, in particular, Phil Layton is the one who designed all the different transgenes from our birds. It's really a very technical exercise, which he's a definite expert. And everyone else in the lab that helped with the injections to create the transgenics, the cell culture, it's really a team effort, because it's a ton of work, [chuckle] and the team at Carterra, Dan and Yasmina could have really helped us over the years in getting all this data, so thank you. At this point, I think we have some time for questions. And again, Noah is also on the line for more technical questions about the LSA instrument. I put our email addresses there if you'd like to email either of us at some point. We're available. Thank you.
0:36:45.2 JM: Thank you, Kathryn. That was an excellent presentation.
0:36:48.2 KC: Thank you, John. [chuckle]
0:36:50.7 JM: We'll now open up for a panel discussion with Kathryn and Noah Ditto. Noah's Carterra's technical product manager. He supported drug discovery and early clinical development for nearly a decade at Bristol Myers Squibb. He's an expert in label-free biosensor screening technology and is one of our longest-term employees. If you have any questions, please send them by private chat to me, John McKinley, the host, and we'll answer them. Kathryn, we have a couple of similar questions that came in, so I'll ask them both to you. Why do you select the kappa lambda light chains? And another one, do you see a significant difference in birds that express human lambda versus kappa?
0:37:41.0 KC: Those are great questions. We selected both of our light chains based again on in vitro data that we gathered. There's a cell line of chicken D cells, DT40, that is really useful for this type of analysis. So we looked at expression in the cell line with our heavy chain to see what expressed the best, and also just the frequency in the population. Jake Glanville has a couple of really good papers out there looking at large populations and the frequency of different V genes in the population that you should look at. As far as kappa versus lambda, this is a question that we get a lot because I think this is changing too, in the antibody discovery field, and most of that work with transgenics has come from mice and mice naturally express kappa at a very much higher rate than they do lambda. So most of the monoclonals that are in the clinic come from mice, and they come... They're usually kappa antibodies. I think that's changing. I think there's more lambda antibodies that have been discovered out there that people are getting a little more familiar with.
0:39:08.0 KC: The chicken only has one light chain locus, and it's more lambda-like than kappa-like, so we made both versions in our birds. And we just published a paper in class one showing that there really is not a difference in immune repertoire between those two birds, in affinity and epitope binning and sequence diversity. So that's nice to see. So we have both available if you're more comfortable with kappa, we have birds expressing kappa, but I think that preference for kappa is more of a historical thing related to mice. Yeah, that's a good question.
0:39:51.8 JM: There's a question about slide 14, data related to sequence variability in wild type chicken or to immuno-dominance by an epitope.
0:40:08.4 KC: Or do we see immuno-dominance? We generate... Usually for each of the transgenes that we express, we generate about 100 clones, kind of a random number that we picked, and we try to look at several different birds. So like for the lambda paper that we just published, I believe we looked at five different birds and we generated 100 clones or 200 clones, in that case. But it's... With that small number of clones from each bird you're bound to pick up dominant clones, and so we do our best. All the clones that we present in our paper, the sequence is unique, but some of them will be related, and I think that's just a function of numbers, so if we picked more we would get... That would sort of be diluted out, if that makes sense.
0:41:10.3 JM: Oh, okay, here's another one. Your lambda light chain has much higher titer in this example, is this commonly observed?
0:41:22.3 KC: Yeah, it's kind of interesting, and we're still doing work to sort of investigate this and we recorded this in the first paper we published with the OmniChicken that the kappa expressing birds, the titer was about 16-fold. If you just look at total IgY, which is chicken IgG, if you look at total IgY, it was about 16-fold less than wild type birds. And we don't see that when you... In lambda-expressing birds, that difference is much less. I think it was like five or eightfold, something like that, I can't remember the exact number. So we do see higher overall titers with the lambda birds, and we're sort of investigating whether or not... We might have some clues into the structure and how that might affect that, but it's a complicated question. We've looked at melting temperature, thinking protein stability might have been it, but there's really no difference in melting temperature. I could point out that our... The Z regions of our birds are chickens but the constant regions are... I'm sorry, the Z regions are human, but the constant regions are chicken.
0:42:43.6 KC: So in the bird, it's a chimeric molecule and you kind of have to do that because you don't wanna interfere with B cell development and cell signaling. So we're sort of investigating whether or not the lambda, human lambda and human kappa a little differently on that constant region and how that might affect expression. But the bottom line is we don't see any difference in epitope coverage or affinity of the molecules that we get out of these birds. Yeah, it's a good question and it's something that we're investigating actively.
0:43:22.8 JM: Thank you, Kathryn. Noah, it looks like we have a question for you.
0:43:28.2 Noah Ditto: Yeah.
0:43:30.2 JM: What are the typical assay formats used for kinetics and binning of bispecifics?
0:43:36.8 ND: Yeah, so that's a good question. They are fairly similar, kind of depends on what your starting material is, so if you had a bispecific that was high percentage or high purity, so if it was with mostly the heterodimer, for example. Katie made good mention of this, with the OmniClic approach that you do want to get the heterodimer as opposed to getting non-ideal forms like the homodimer. So if you do start with that purity level, the assays can be run as they would for traditional IgG monoclonal antibody, where we would for kinetics, you would, for example, capture via the FC region and run the antigen as a titration across that to do kinetic modeling. And then in epitope binning, the assays would be similar as well, we would do a direct coupling typically to the LSA surface and then flow your antigen and antibodies across to ascertain competition events. So it really comes down to the type of molecule being offered. If there is a mixture, there are more complicated assay formats like bridging-type assay formats, where you would first capture with one antigen followed... And then come in with the second antigen with the idea that if you did detect binding with the second antigen, that might give you indications that there was a mixed population and it rules out the complications that having homodimeric forms, for example, can introduce.
0:45:06.1 ND: So ideally, really, for assay simplification, it's great to start with what like Katie described here, again, with OmniClic and having a high purity, heterodimer population to start with, it makes all the assays much more straightforward.
0:45:19.9 JM: Here's another question for you, Noah. How much sample is needed for kinetics and binning assays on the LSA?
0:45:29.1 ND: Yeah, so in terms of, for a kinetic assay, assuming we do the capture approach that I just mentioned, you're looking at maybe a few micrograms at most of each clone, that's probably on the high side even. It could probably be a little bit lower than that. And then for binning, it kind of depends on the assay format and the types of signals you get, but you could assume around five micrograms might be needed. So collectively, to do a kinetic and binning analysis, you're typically well under 10 micrograms total of each clone needed to carry out both exercises.
0:46:03.8 KC: We do not purify any of ours. I mean, we have but most of the assays, most of the data that you saw in the talk was with unpurified supernatant, it's really handy.
0:46:18.2 ND: Yeah, that's a great point too. So in these assay formats, like if you did a capture format, just like Katie mentioned, the instrument, the fluidics of the LSA are such that you actually do on censor ChIP enrichment based on our flow dynamics, so that you can start with a crude matrices and effectively reach what you wanna actually measure. And as Katie said, it's usually straightforward and there's minimal... There really is no up front work to be done prior to the analysis.
0:46:48.1 JM: Thank you, Noah and Katie. Oh, Kathryn, sorry, for the... Kathryn, do you observe any difference in the number of B cell antigen-specific between OmniChicken and OmniClic?
0:47:05.4 KC: Anything specific. It's not really something that we measure. In the programs that we've done with OmniClic, and certainly the progranulin program, in the comparison between Clic and OmniChicken, we've seen similar... If you look at those graphs of changes from germline, we've seen no statistical difference in that number between the regular OmniChicken and OmniClic, if that answers your question. I think actual numbers of B cells is a little hard to answer, but if you look at the downstream metrics versus the affinity, we don't see a difference.
0:48:04.9 JM: Okay, about the monoclonal antibodies that are diverse in primary sequences but bind the same epitope, the purple P domains on the antigen, is the sequence diversity clustered at the CDR regions?
0:48:23.8 KC: Yes, definitely. The way that we count diversity or diversity in our clones is by looking at CDRs, so yeah, different CDRs. We don't allow... I know some people will count clonal families, and they'll say one or two amino acids differing in CDR-3, that's not the way we do it. Every change is a different molecule.
0:48:50.6 JM: Okay. Even though the VL/VH in the OmniChicken/OmniClic platform are from human, are there any developability issues that you've observed later in research or developmental stages?
0:49:10.1 KC: Yeah, that's a great question. We get a lot of... Early on when we introduced the chicken we were asked a lot of questions about the humanness of those sequences, and so we did do a project with Effevax, and some other groups looking at the LakePharma T20 score also, and we scored very human. There was no difference between our human sequence antibody clones and a panel of randomly selected human clones from the database, so they are quite human. We're working right now with a partner looking at sequence liabilities in our clones, and I can say that that data looks really promising, and so hopefully we'll have a paper coming out soon describing that and also publishing on some of that Effevax data. But yeah, we're really happy with the degree of humanness of our clones and sort of the minimal sequence liability that we've seen so far in the data.
0:50:20.3 JM: Thank you, Kathryn. Do you have the ability to induce immune response against whole cells with your platform or do you need purified epitopes or proteins?
0:50:32.3 KC: Yes, we have done cells... The challenge with that is having something to screen on, so we have immunized with whole cells, but you really have to be pretty diligent in what you're using as your screening material, 'cause you're obviously gonna mount an immune response to other proteins on the surface of that cell so you really have to have a really good reagent to screen with, but yeah, we have done that. We also do... We can do DNA immunizations, either with the gene gun, the Bio-Rad gene gun or we do a 293fectin injection. Recently, we acquired a company specializes in transmembrane proteins, so we're sort of building the capability now of being able to express transmembrane proteins in vesicles with different strategies and generating our own reagents for immunizations, so if you do a program with us, I believe that's one of the options to use that service. It's a really great team of scientists that can help you with your transmembrane protein expression.
0:51:54.3 JM: In your transgenic chicken, do you see more epitope diversity than mouse for the same antigen?
0:52:01.8 KC: Yeah, that's a great question. So that goes back to the very first paper that we published on chickens. In the first paper, we compared wild type... A panel of wild type antibodies, we got progranulin against the panel of mouse antibodies that was done by another company at a different time, and we saw really the same epitope coverage at the mouse. But we also saw a couple of different epitope bins that were not seen in the mouse. It's hard to say if those were really unique or not, because the panels were so small, I suspect that if you probed the mouse a little more, you might see the same thing, but it was interesting that we pulled those up right away. So then what we did is we took the panel of antibodies from our wild type bird, and those are what I was referring to as our standard antibodies. So we knew what epitope bins those represented and we've used them subsequently in every experiment with our transgenic animals as a comparison. And in our comparison with the transgenics, we hit every epitope bin that those wild type birds have hit, and in turn those have hit every epitope bin that the mice have hit. So yeah, we see very similar epitope coverage. I believe we presented a few years ago in a conference, there was another antigen that we worked on with a partner that we in fact did pull up several different bins that the mouse platform had not. So that was probably nice to see, and I think that was presented in a talk at DOT or something like that.
0:53:47.3 JM: Thank you, Kathryn. It looks like this question is for both Kathryn and Noah. On the Carterra LSA, how long does it take from concept to lead?
0:54:06.9 ND: Do you wanna take a stab at that Katie or?
0:54:10.7 KC: Well, I can speak to the immunization [chuckle] aspect of it. Once we get an antigen from a partner, the immunization is usually six to eight weeks. And then the screening depending on what kind of antigen you're working with, is it cells, vesicles, or just soluble protein, that... But I would say, at least for the progranulin protein, it's a few weeks before we have a good number, maybe like a month before we have a panel of antibodies that we can run. And then as I said, we just send the crude supernatant over to the instrument.
0:54:58.7 ND: Yeah, and if the question also included the time component for the analysis for kinetics and binning, typically, it's to screen 384 clone affinities, up to 384 clone affinities. It's about maybe a 12-hour of total runtime, maybe a little less depending on how you set up the experiment. So in a long day, or an overnight run, you can generate 384 of them. And on the binning side, it kind of just depends on how many clones you have in the assay. If you did have a full 384 panel, it would take closer maybe to a week to run, if that panel was smaller and say maybe 96 clones or something like that, it would be on the order of 24 hours or so. So that just is a ballpark of experiment runtimes typically.
0:55:50.5 JM: Thank you, Kathryn and Noah. It looks like we have the time for one last question. Kathryn, what is the rationale for just using a single VH gene for the transgenic chicken, instead of using diversed human VHs?
0:56:09.1 KC: Yeah, that's a good question because most of the mouse platforms out there will include, including our own, include the entire complement of V, D and J genes. Part of it is technical. The chicken itself, the wild type chicken only has a single V gene. So it does rearrangement with a single V, single D, single J, and a single V and J on the light chain. So that's the technical aspect of it. We, I think in our group, we looked on that as a bonus, VH3-23, for example, is really highly expressed in the human population, much more significantly than its closest runner up, I don't remember the exact numbers. But it's very highly expressed in the normal human population, and it also, it simplifies the downstream process to not have to do primer cocktail, which when you have 15 or 20 primers in a PCR reaction, it's really cutting down on your efficiency and you're probably losing clones. So when we do our single-cell RT-PCR, it's four primers in there, we get great amplification. So yeah we've done our best to try and choose V genes that are common in the population. And I think it's worked out pretty well. It's a really simplified process downstream, and we don't see... Again, we mentioned that comparison of the wild type to the transgenic mouse. We didn't really see any difference there in immuno-competence.
0:57:51.8 JM: Well, thank you, Kathryn, for such an informative presentation. And we really appreciate it.
0:57:58.8 KC: It's all good, yeah.
0:58:03.2 JM: And for those of you whose questions we didn't get a chance to answer, we'll make sure to reach out to you after today's webinar. And also, if you have any other questions for Kathryn or Noah, please send them an email and we'll make sure your questions get answered. Thank you all and we really appreciate you attending today's webinar. I hope you have a great day.
0:58:25.7 KC: Thanks, John.
0:58:28.4 JM: Thanks, Kathryn.
0:58:29.4 ND: Thank you, everybody.