Accelerating Antibody Screening with Array-Based Surface Plasmon Resonance – January 17, 2018

0:00:00.7 Jeff Buguliskis: Hello everyone, welcome to another Genetic Engineering & Biotechnology News webinar. Our presentation today is entitled Accelerating Antibody Screening With a Ray-Based Surface Plasmon Resonance. I'm Jeff Buguliskis, Technical Editor for GEN, and I'll be the moderator for today's webinar presentation. Commercialization of biologic antibodies has been on an exponential climb in the last several years, and due to extremely successful products for a multitude of diseases, shows no signs of slowing down. A greater understanding of binding kinetics and epitope coverage of monoclonal antibody libraries is essential for guiding the development of next generation therapeutics. With the relatively recent emergence of surface plasmon resonance or SPR arrays, investigators have the high-throughput capability to accelerate their bio-therapeutic monoclonal antibody screening campaigns while using minimal volumes of sample that are often available only at low concentration and limiting quantities, thus eliminating the notoriously tedious and resource-consuming endeavor of collecting this type of data.

 

0:01:09.6 JB: Well, let's meet our speakers for this webinar who provide data from recent case studies that screen antigens over large monoclonal antibody panels, as well as a comprehensive epitope-binning analysis utilizing SPR array methodologies. Kathryn Ching is a Senior Scientist at Ligand Pharmaceuticals. Dr. Ching will describe how Carterra's array SPR platform heated her epitope mapping endeavors from transgenic chickens. Yasmina Abdiche is Chief Scientific Officer of Carterra. Dr. Abdiche will provide an in-depth specific of SPR technology and how it is utilized to generate high-throughput kinetic data for epitope mapping and how Carterra's LSA device using SPR technology will facilitate researchers antibody screening projects. Before or speakers get started, I wanna encourage the audience to submit questions for our Q and A session at the end of presentations. We'll try and answer as many questions as we can. So simply type your question into the Ask a Question area below the presentation screen and hit submit. Alright, well, let's get our webinar underway. Kathryn, we are listening.

 

Abouth Kathry Ching

0:02:22.6 Kathryn Ching: Good morning, I'm Katie Ching from Ligand Pharmaceuticals in Emeryville, California, and I'd like to thank Yasmina and her team at Carterra for inviting me to present at this webinar in this morning. We've done a lot of work with Yasmina in the past, and we really appreciate the expertise and advice that she and her team have given us over the years to better understand our transgenic animals. You've probably heard of Ligand through the OmniAb platform for antibody discovery. We have the OmniRat and the OmniMouse carrying human B genes. Also the OmniFlic for bispecific antibody discovery, and now we've added to that platform, the OmniChicken, which is what I'll be talking about today.

 

Why Chciken?

0:03:09.8 KC: The first question that chicken people always get asked is, why chicken? And the answer to that is evolutionary distance. We last shared a common ancestor with mice about 95 million years ago, and so that means that a lot of really important proteins that are targets of drug discovery today, such as cell surface receptors, cytokines, cell signaling molecules, a lot of these proteins are really highly conserved between mouse and human. So when you immunize a mouse with that human protein you're interested in, there's a really limited epitope space available to that mouse, and that's gonna be reflected in the panel of antibodies that you generate from those animals. In contrast, we separated from chickens about 300 million years ago, so that same ortholog in a chicken looks very different than it does in the mouse, it's gonna be much more immunogenic, you'll have broader epitope coverage of the molecule, and you'll probably likely have much higher affinity antibodies, because of that greater immunogenicity. So the OmniChicken is actually several different lines of transgenic birds in which we've replaced both the light and heavy immunoglobulin locus in the chicken with human B genes, when we have several different flavors of each white and heavy chain in individual birds.

 

0:04:31.7 KC: So what I'm showing you here is the schematic of our heavy chain or the two different heavy chains that we have available. And if you're not familiar with chicken immunology, it's a little bit different than humans and mice. Chickens actually rearrange a single of functional V, And then they generate diversity through a process called gene conversion with pseudogenes upstream. They do do VDJ rearrangement, but they don't actually get a lot of sequence diversity from this rearrangement. So what I'm showing you here on the top line is or what we call our SynVH-C gene, and this is a pre-rearranged human VH3-23 and it's downstream of an array of human pseudogenes that we've derived from human ESTs. And of course, we've kept the chicken constant region, so that cell signaling and all those interactions can go on as normal in the bird. The other version that we have of the heavy chain is our rearranging heavy chain, which we call SynVH-SD and that's that second line. So this gene can actually do VDJ recombination, it's again VH3-23, it's downstream of an array of human pseudogenes and carries the chicken constant region.

 

SynVH-C Chain

0:05:49.7 KC: We actually published just recently a paper describing the SynVH-C heavy chain paired with a kappa light chain, and if you're interested in a lot more of the technical detail and how the knock outs and knock ins were generated and also more of the cell biology and characterization of the bird that is available in mAbs and it's open access. What I wanna talk today about is our SynVH-SD bird, and some of the characterization that we've done with Yasmina in learning more about these birds. The first thing that we always look at in our birds is what their B cell population and T cell population looks like in the periphery. So this is cell surface staining in a number of wild type and SynVH-SD birds and B-1 on the far left is a marker for chicken B cells, and you can see that there is no difference between the control and SynVH-SD birds. The same with cell surface IgM and IgL, and also two different T cell markers at the end of the graph. The VH3-23 V-kappa 315 in the center of the graph is actually an antibody that we generated against our germ line insertions. And what you can see very nicely is that the control birds, the wild type birds don't show any staining with this antibody, but the SynVH-SD birds do, verifying that these chickens are expressing human antibody at the cell surface.

 

0:07:24.3 KC: When we immunize these birds, we often see very robust titers. On the left, you can see a chicken that has been immunized with a protein called progranulin, which I'll be talking about in a bit, and you can see right away that at draw 2, we can still see signal at a 1-10,000 dilution of the sera from the bird, so very robust titer came up very quickly towards this protein. On the right, I put this in because the titer is not as robust. You can see that after four boosts, we're seeing a very... Not a strong signal at about 1-1000 towards the target, but I put it in because we did generate antibodies to this target from this bird, and I think it shows that our process is such that even with such a low titer, we're still able to pull out specific antibodies.

 

What we will do with all our birds?

0:08:20.0 KC: What we really wanna do with all our birds is be able to compare between genotypes, be able to see what the different transgenes, how they perform and how they compare to one another. So what we've done to do this is we use what we call our model antigen, and that is the human progranulin protein. And we've compared immunizations in different transgenes using this antigen in a number of different ways. Progranulin is highly immunogenic in both mice and chickens, and we actually have a panel of mouse antibodies towards this protein that we've shown in our wild type birds, we get the same epitope coverage as what you see in a mouse. And that paper that I showed you earlier, is actually showing that our SynVH-C kappa birds also has the same epitope coverage with the wild type birds and as the mouse. But in addition, our antibodies are cross-reactive to mouse, so that's very helpful in your drug development process to have the species cross-reactivity.

 

0:09:29.4 KC: We immunized a small cohort of SynVH-SD birds with the progranulin protein, and we use our GEM technology to generate a small panel of about 75 antibodies to look at their different characteristics. So if you're not familiar with our GEM technology, it's a technology that allows us to identify up-front B cells that are secreting antigen-specific antibody. We then take these individual be cells and we're able to do single cell RT-PCR and isolate those paired heavy and light chains. Isolate those paired heavy and light chains, clone the VH and the VL into a single chain Fv-Fc vector and express those as expressors in a mammalian cell line, and then we can do all manner of different downstream assays with that crude supernatant. So what we did was we generated this panel and we sent these crude supernatants to Carterra, and they did a number of different assays to look at the... To basically analyze the epitope coverage that we had with this small cohort of birds. So they did a basic sandwich assay in which they looked at different pairings of our antibody panel using their instruments, so how different antibodies can cross-block or not cross-block one another. And they also did an assay in which the antibodies were used as analyte and flowed over a chip that's coded with either the full-length progranulin or different sub-domains of the molecule, so that we can better define the epitope, the epitope bin that an antibody might fall into.

 

0:11:23.2 KC: And in both of these assays, they run what we call our wild type standards, so these are highly characterized antibodies that we generated in our wild type chickens, and we know what epitope bin they fall into, and we use those as sort of markers at the different epitope bins. So what you're looking at is nodal plot of that data from three different SynVH-SD birds. And what you can see quite clearly, the different bins are color coded, and I don't think I mentioned earlier, but the progranulin protein is a seven... Has seven different sub-domains, so it's really amenable to epitope mapping. And what you can see very clearly is that we hit all seven epitope bins in the progranulin molecule with this very small panel of antibodies from the SynVH-SD bird. What I think is really interesting also about this kind of plot, as you can see the different interactions between antibodies and whether or not they cross-block one another within an epitope bin. You can see their relationships to one another.

 

0:12:30.3 KC: Finally, we also like to look at the sequence data of our antibodies in comparison and compare it to the epitope binning data that we get from Carterra. And so what you're looking at on the right is a sequence dendrogram, so looking at how these different sequences are related to one another, stacked up against the epitope bin that has been identified that each antibody hits. And what you see... What you expect is similar sequences fall into similar epitope bins, you would expect that. But what you occasionally see is different sequences converging on the same epitope bin, and you can see that here with the P domain, this bird, 29406 hits it right here in the sequence dendrogram and down here, and you can see that they are highly unrelated sequences. So that's kind of neat that you can see that with this data.

 

0:13:30.8 KC: Also, as I mentioned, with the chicken, you get mouse... You can get mouse cross-reactive antibodies and about half of those... Half the antibodies in this cohort were a cross-reactive with mouse. And I'll talk about affinity measurements. Another thing that we like to look at in terms of sequence is where is the diversity generated. So is it generated in the CDRs? In the framework? And from a manufacturing standpoint, you want that diversity to be generated mainly in the CDRs and not in your framework. And that's exactly what we see with our SynVH-SD bird. So this is a graph of the heavy chains of all the antibodies that I just showed you on the previous slide, and what we've done is we've mapped every change at every position in the heavy chain from this panel of antibody. So each different color is a different amino acid change at that particular position. And you can see quite easily that the diversity is coming in the CDRs with just a few changes in the framework.

 

0:14:40.9 KC: Another thing that I mentioned in terms of chicken immunology is that they don't generate a lot of sequence diversity from VDJ rearrangement of... To generate the functional V, but what they do generate is length diversity. And that was one of the reasons why we developed this bird to see if we could get more diversity in CDR length, and we do indeed. That's exactly what you see. So that graph on the bottom is showing that we get anywhere from 8-20 amino acids in CDR-3 from the small cohort of antibodies. Lastly, the data that we get from Yasmina is kinetic data. And this graph is a little bit misleading because we run... So in red, you can see the human sequence antibodies, and in blue, these are our wild type standards. And the wild type standards are a curated set, so they're the best antibodies that we've chosen from I think a campaign of more than 100 antibodies, so they're very high affinity, and we use them as markers in these assays because of that. But in this particular graph that you're seeing, they were also running duplicate, so it's a little misleading in terms of the chicken, the wild type chicken antibodies that are represented here, but I think the take-home message from this graph is that we get high affinity antibodies, antibodies and the picomolar range from our birds, and we see that just even in this small group of antibodies.

 

Conclusion

0:16:16.7 KC: So in conclusion, I've shown you that our SynVH-SD OmniChickens have a normal B cell compartment and normal T cell compartment. I showed you some data showing that they respond quite well to a variety of antigens. I showed you our standard analysis of a panel of antibodies to our progranulin model antigen, and showed you that our SynVH-SD birds demonstrate the same epitope coverage as our wild type birds and as a panel of mouse antibodies. But in addition, again, our chickens have species cross-reactivity because of... They have species cross-reactivity. And I also showed you some kinetic data showing that we can get high affinity antibodies. And that our antibodies, I think really importantly, the diversity is focused on the CDRs and not on the framework. And finally, a bit of data showing that we do, in our rearranging bird, we are able to get CDR heavy chain length of diversity.

 

0:17:27.0 KC: I wanna thank you for your time today and also acknowledge the people at Ligand and who have contributed to this project. There's a lot of work that goes in in embryology and molecular biology before we can even generate a panel of antibodies to look at and we have a really good team at Ligand. And also finally, Carterra, Yasmina and Dan have been really great to work with and really patient with us and just great, great advisors. So thank you very much.

 

0:18:01.8 JB: Thanks Kathryn, that was a wonderful way to begin our webinar, and I think our audience now has a better understanding of the advantages of transgenic chicken antibodies. Before we move on to the next presentation, I wanna remind the audience once again, to submit questions for a Q and A session right at the end of the next talk. We'll try to get to as many questions as we can, so simply type your question into the area below the presentation screen and hit "Okay". Alright, with all that said, Yasmina, the floor is yours.

 

About Yasmina Abdiche

0:18:33.0 Yasmina Abdiche: Hello everyone, thanks for joining us today. My name is Yasmina Abdiche, and I'm CSO of Carterra, and I'm delighted to be here today to tell you about how you can use our LSA platform that uses array-based surface plasmon resonance to accelerate your antibody screening campaigns. I'd like to start with an analogy that is really at the crux of where we see the utility of our technology. So this is a picture of the night sky, and NASA focused a very sophisticated telescope into a part of the night sky that was previously thought to contain nothing, it was basically black. And after focusing this telescope for many, many hours, and then they collated the pictures, they realized that there were millions of galaxies there, that they hadn't known about before. And the analogy here to our platform is that when you have a higher resolution and a more sophisticated tool, you can see more things and therefore be more informed about the antibodies that you have already. So I'm really seeing a paradigm shift in the industry now that people are screening more what they have rather than doing a lot of engineering afterwards.

 

Antibody Therapeutics

0:20:00.4 YA: So antibody therapeutics are really a growing and lucrative part of the pharmaceutical industry. Primarily it has in the past been dominated by small molecules, but one of the main blockbuster drugs is actually an antibody therapeutic, HUMIRA, bringing in about $16 billion a year in sales. And so there's been a shift in the analytical tools that people are demanding to explore interaction analysis using different modalities, not just small molecules, but antibodies also. And biophysics really plays a key role in drug discovery because it's really agnostic of the molecule's modality or the therapeutic area, because it measures molecular level binding events. And it's really important to drug discovery to understand how your molecules work, so mechanism of action is really high on the agenda, and biophysics can really help with that.

 

0:21:05.8 YA: So on to Carterra's solution, our LSA, which stands for Lodestar Array, integrates flow printing and array SPR into one instrument. This really allows you to print out a large panel of proteins, up to 384 per array, and then that flow cell, that multi channel flow cell will undock and then what will dock over your chip is a single flow cell. And then you can do interaction analysis across the entire array in a single step. So this instrument allows for automated flow cell switching between these multi channel and single channel modes. And you can do inline reloading of an array, say, for capture kinetic experiments. And the chip chemistry support all the standard formats, such as amine coupling or capture via a reagent such as anti-human Fc. Just a little visual on what the array would look like, this is a cartoon showing how 96 spots are created via the docking of the 96-channel print head, and by undocking and redocking you can nest up to four prints to create a 384 array. And then by switching to the single flow cell geometry, you can flow your analyte across the entire array, and so get one analyte binding into 96 ligands or one analyte binding to 384. And it's really up to you how many ligands you put down on the surface. So this is exceptionally efficient on analyte consumption because one injection gives you data for multiple reaction surfaces in a single step.

 

0:23:03.1 YA: I'm going to review three main applications that are really relevant to antibody discovery in a therapeutic setting. The first is going to be binding kinetics and affinity. The second will be epitope mapping, which really only is relevant for linear epitopes. And since most antibodies recognize confirmation epitopes, I'm gonna spend much more time on epitope binning, which is a pairwise combinatorial competition assay. And there are some new publications that we've had where we demonstrated how array SPR is particularly well-suited for epitope binning applications, and I've just listed them here for you. So for kinetics, this is an experiment where 384 antibodies were captured onto an anti-human epsilon, and then a single antigen was injected across the entire surface at titrating concentrations. And so the data shown here shows examples of three different clones with very different kinetic profiles, showing that you can have a wide dynamic range accessed in a single experiment from tighter than 80 picomolar to, say, half micromolar in the KD value. And I've also shown the data from a blank spot where no interaction was observed, and this just gives you confidence that there is no bleeding or cross-talk between the spots.

 

 

0:24:44.0 YA: Another really important feature I'd like you to notice on the slide is that the data are being fit to a simple model. So the simple model is shown as the red lines and the blue lines are the measure data. So the data quality is really high, so you're not sacrificing the confidence in the data by doing this experiment in high-throughput. And because the analyte sees the entire surface, you need a very small amount of it, and that means that you not only save sample, but the sample set up is facile also. So what does high-throughput kinetics look like? So this is a snapshot from our dedicated kinetic analysis software, and it shows you the output of a typical experiment run on a 384 array. So each individual panel or a stamp in the stamp collection is an entire binding profile of the antigen binding to that spot that has an antibody immobilized in it. And our software has inbuilt into it flags to call out poor quality data, and this is really important when you're generating large volumes of data. So the gray panels show spots where no binding or barely binding responses were observed. Those may be intentionally blank spots, inactive antibodies or antibodies that were captured at a very, very low capacity. And then the yellows, they indicate spots where the data is heterogeneous and therefore the fitting function is not valid.

 

0:26:29.9 YA: We also have included a provision for not enough curvature in the on rate, which is shown by the purple panels, and this really allows you to weed out those poor binders so that you are only focusing on high quality data, which really can give you confident binding kinetics. So basically, what I just said is that the software allows you to highlight the good, the bad and the ugly. Another cool thing about the software is that we can automatically generate these ISO affinity plots, and this really gives you a great overall description of where the kinetic rate constants lie. You may have similar affinity clones that are reached through different kinetics, and this allows you then to understand the kinetic diversity of the panel. And so in a single experiment, you can see here, there's orders of magnitude different binders in terms of affinity. And so this is really a very important at early stage characterization, you can characterize a really broad range of affinities.

 

Epitope Characterization

0:27:44.0 YA: I'm gonna switch now to epitope characterization, and as I mentioned, knowing the epitope or knowing how your antibody functions is really important in drug discovery of antibodies for therapeutics. An antibody's epitope largely dictates its function, so if you saw antibodies by epitope is more relevant functionally than affinity ranking them. Another thing is that an antibody is sort of born with its epitope. It's innate, it can't be shifted rationally by engineering, and you can't really predict it or design it in silico. So therefore you have to select the epitope that you want. Often times you don't even know what epitope you want, so therefore characterizing what you have is a really important step in influencing the decisions that you make. Identifying antibodies with unique epitopes is really important for securing IP and also to differentiate your product in some way from other products that may operate by a similar binding mechanism. So therefore, if you can survey the epitope landscape of your antibody panel at the earlier stages of research, there's huge benefits to you in honing in very quickly on interesting clones that you can prioritize your resources over. So by essence, assays can really provide you powerful tools to allow you to sort antibodies into epitope families using either mapping or binding methods.

 

0:29:28.2 YA: So first, I'm gonna talk about epitome mapping. And for this example, I'm going to show some data that was published several years ago, showing the application of biotin-related peptides arrayed on different commercially available biosensors, and then the throughput of experiments where antibodies are screened over those arrays. And so in this particular experiment, there was a set of 24 overlapping peptides that were arrayed onto the surface and then a bunch of antibodies that were screened over them. While all the biosensors could give the same results, the same mapping results, there was definitely a trade-off in terms of time taken, amount of sample used, and the kind of complicated setup of the assay needed. And now we shift forward to array-based peptide mapping. This has huge advantages in throughput and in minimal sample consumption. So in this experiment, library of biotin-related peptides was arrayed on to individual spots of sensor chip, a streptavidin-coated sensor chip, and then that was used to screen a large number of antibodies, and this really uses a very tiny amount of sample.

 

0:30:53.8 YA: This is the data analysis from our dedicated epitope tool software. This one-on-many format really expedite this kind of application. So what you're looking at on the left panel are some sensorgrams from the analysis, and you are able to set a threshold cut off of where binding occurs versus no binding or residual binding from background. And then a heat map is automatically generated and sorted. We also generate dendrogram and stacked plots, so you can really navigate the data very, very quickly and cost your antibodies into epitope families.

 

0:31:36.8 YA: This is the summary of the data for a simple overnight experiment where we injected 96 antibodies one after another, over a 384 peptide array. Since we had few than 96 antibodies, we had 22 antibodies injected in replicate cycles, and because we had fewer than 384 peptides, we arrayed our library multiple times onto different spots of the array. So what you can see here are blocks of data that show that spot to spot and injection to injection are reproducible. And so very quickly we were able to cluster those antibodies into different mapped families as shown by the boxes, at the bottom of the slide. In this particular experiment, the peptides were a library against the mouse target and the human target. We very quickly were able to identify binders that were mouse specific, human specific or cross-reactive between the two. We also identified some binders that bind into the soluble target were binders but did not bind to peptides, so those we would conclude by a confirmational epitope that can't be recapitulated on the surface of... On peptides.

 

0:33:00.4 YA: So moving on to epitope binning assays, because most antibodies are likely to bind a conformational epitope and their epitopes may not be able to be re-capitulated on peptides and therefore would not be amenable mapping assays, you can use competition assays to cluster antibodies into epitope bins or epitope families. So there's various ways that this kind of experiment can be oriented in the context of a biosensor. You can do an in-tandem, pre-mix, or classical sandwich format. Each one of these varies in terms of which antibody or antigen is immobilized onto the surface and whether the antigen sees the first antibody in solution or on the sensor. So this assay scales geometrically with the size of the antibody's antibody panel, and so any way to make this more efficient and actually use the two-dimensional array format really does expedite this analysis. So because binning assays scale geometrically, historically, they've been really limited in size to really a handful of antibodies, and I just show an example here of really nice paper by Mousa et al., published in 2016, so relatively recently. But they took a relatively small panel of only 13 antibodies and did an epitope binning experiment.

 

Array Antibody

0:34:33.9 YA: In the same year we published a paper using array SPR, where we demonstrated epitope binning on a 384 array. And here I show the data for 350 antibodies cross-competed against one another. And the amount of data that you can generate is just massive compared to traditional methods. So this is a snapshot of what the data would look like coming off the array biosensor. You have a bunch of cycles, and the left-hand plot is what I like to call a Mountain View plot, where each cycle has been concatenated to the next, so you can see where sandwich signals are occurring or where there is blockade. And then on the right-hand side is an overlay plot of the data from a single spot coupled with an antibody. So this is taken from our software, and you can see that there's various thresholds that the user can apply to determine whether you have a sandwiching single... Signal or no signal. And using the data from this kind of threshold setting, a sandwich signal would be a green, like a traffic light, and a blocking signal or no signal would be red.

 

0:35:49.1 YA: And so that information is put together into a heat map, and so shown here is a published example where we did antibody array of 192 antibodies. The heat map on the left summarizes all of the epitope cluster information where the coupled antibodies are shown in rows and the solution antibodies are shown as columns. Then the same information is captured visually in a network blocking plot. This shows the different communities where the antibodies are hanging out. And we like to kind of think of this as like the Facebook of antibodies, it shows you which antibodies are clustering with which other antibodies, where the relationships, the blocking relationships are occurring.

 

0:36:38.6 YA: So our software will automatically generate the sorted heat map and the network blocking plot. And we take that further by allowing users to enter in other information into the epitope tool. For example, if you know that your antibodies cross-react to mouse, you can list your antibodies with either "Yes" or "No" to mouse cross-reaction, and then color code the network blocking plot by that information. Similarly, if your antibodies are derived from different libraries, you can use that information to color code the network blocking plot. And by merging data from orthogonal sources, we find that the network blocking plot is really informative to consolidate a lot of information and allow you to identify antibodies with unique properties that you may want to pursue further. You can read more about this experiment in a recent paper by Ching et al. In mAbs.

 

0:37:43.0 YA: So another example I'd like to discuss is this antibody binning experiment that we did several years ago on a 70 antibody panel. Similarly, we generated the heat map, we generated a lot of data, which we then merged with other orthogonal data. I wanted to draw your attention to another profile that we can see in the sensorgrams. So in addition to an analyte being blocked or not blocked by binding to the antigen that is first captured by the antibody on the sensor surface, we also see this other behavior where you have an inverted sandwiching signal highlighted in yellow here called displaced. And I'm gonna talk about that at the end of this talk. So similar to the other paper that I described, when you merge in orthogonal data, you can make the network plots really informative. So what we noticed here is in a functional cell-based assay, two of the bins, which we discrete from one another, we're both functional blockers. So this was very interesting to us because it would suggest that two different types of epitopes are converging upon the phenotype of blockade. And this experiment was done on antibody sourced from healthy individuals, and we noticed that each of those things were populated by antibodies from all of the donors, so it seemed to be kind of a universal mechanism that were shared by nature.

 

0:39:25.6 YA: By looking at their sequences further, we also showed that each of the bins had a very strong bias toward a particular germ line in the IgG heavy chain. And then when we did some mutagenesis mapping, we noticed that each of those two blocking bins were actually targeting a different part of the protein, NEAT1 and NEAT2 in this example of IsdB. And so when you put all this information together, you can see that the network blocking plot can really give you a holistic way of viewing a multiparameter way of looking at your data. So the last thing I want to discuss here is antibody displacement, so this is a really cool phenomenon that you can easily identify in an epitope binning experiment that is conducted in a classical sandwich format. So as I mentioned before, when you have a solution antibody and you're looking for pairing with an immobilized antibody, you usually have one of two outcomes, either the analyte does not bind and you infer is being blocked, or the analyte does bind and you infer that it's sandwiching because it's binding a non-overlapping epitope. But sometimes there's a third category which is kind of a hybrid of the two, that the solution antibody and the immobilized antibody can both bind the antigen but form a transient complex that actually rearranges to displace the ligand and then the analyte leaves with the antigen, and this is the cause of that inverted binning signal that I mentioned earlier.

 

0:41:10.1 YA: So using array SPR, you can see this very clearly in this sensorgram plot, so in blue are buffer analytes which do not show any sandwiching signal, and this shows the natural decay of the antigen from the immobilized antibody. And the inverted signal there indicate a displacement is happening. So this information can also be incorporated into the community or network plots. We use a dotted line to show an asymmetric block where Antibody A blocks Antibody B, if A goes first, but not if B goes first. So these plots can be highly informative. A neat thing about antibody displacement is that we found that two antibodies with disparate affinities through their off rate, so 1004 different affinities, both have the same potency in displacing other antibodies. This suggests that the phenomenon is concentration-dependent and you can drive it to completion through concentration. So it's not just that low affinity antibody is being displaced by high affinity antibody. Those two antibodies can just place another antibody if they have the same on rate, or if you can compensate for the effective concentration by pumping up the concentration.

 

0:42:40.5 YA: To really understand when a displacement occurs from a structural perspective, we did an experiment using antibodies from EGFR, and these four antibodies that bind very close epitopes and are all available through patterns. We made these antibodies and did a pair-wise competition between them. And we showed that some antibodies block one another, some sandwich and some displaced one another. Because we knew the crystal structures, we could then understand the mechanism behind the displacement. What we found is that antibodies with closely adjacent or minimally overlapping epitopes can displace one another. This is shown here very nicely with these space-build models. So two antibodies that clash will block one another, two antibodies with no shared contacts will sandwich pair, but ones that are kind of in between that may not overlap at all, but are very closely adjacent may this place one another. This is a cartoon, just pictorially showing what is happening. So for example, in red, we have one antibody on the surface of the biosensor, it binds EGFR, and then in a sandwiching experiment, we titrate in another antibody. A transient complex is formed, but the first antibody is being kicked out as we raise the concentration of the analyte antibody. And then what happens is that the free ligand is then regenerated at the end of the experiment.

 

0:44:28.5 YA: So I hope I've have allowed you to see how biosensors are very versatile in drug discovery, especially of antibody therapeutics, because they really do provide a label-free real-time way of monitoring molecular level interactions. And they can be applied for kinetics, affinity, epitope and specificity, which are all critical parameters in understanding mechanism of action in guiding drug design. And our platform really takes throughput orders of magnitude higher than what is available on other biosensors. And you get a very efficient use of your samples through this highly parallel analysis. Because the sample set up is very easy as well, you're saving time there as well. And we have sophisticated and dedicated software to enable the analysis to be easy. And I'd like to give some acknowledgements here for some of our collaborators, and thank you so much for spending time with us today.

 

0:45:43.2 JB: Thanks Yasmina, for the very informative talk. I'm sure our audience now has a much greater appreciation for the importance of accurate high-throughput antibody screening technology, and for the versatility of the LSA platform. Before we start the Q and A session, I wanna let everyone know that is their final chance with their questions for our speakers, so hurry up and send them in now. Alright, it looks like we have a bunch of really great questions that have already come in, so... Well, let's begin the Q and A, and we'll try to get to as many as we can. Give us a few moments on our side to get all the logistics squared away, and we will begin the Q and A presently.

 

Question

0:46:23.5 JB: Alright, everyone. So let's begin the Q and A. Our first question is gonna be for Katie. Katie, one of the audience members would like to know, can you give an example of a target that failed in mice, but succeeded in chicken?

 

0:46:43.6 KC: Yes, we actually have worked on a target called BDNF, that's a target in a lot of neurology indications, and you actually can't get even a titer in mice because of the high homology between the mice on human ortholog, I think it's about 96%. And it's actually still quite high in chicken, I think about 91% or 92% homology between the two species. But we immunized two birds with BDNF and we saw a titer after the first boost, and we were able to generate a good sized panel of antibodies to the molecule. And actually I think we talked a little bit about it in that mAbs paper that I mentioned. And Yasmina also did some epitope binning on the molecule. So yeah, that was a great example where you can't even get a titer in mice, but we see it in chicken.

 

Question

0:47:37.8 JB: Thank you very much. Yasmina, our next question is for you, one of our audience members would like to know does Carterra offer collaboration projects with academia.

 

0:47:47.4 YA: Yes, absolutely. Just email me. We'd be delighted to collaborate on any project that is maybe towards a co-authored paper or if there's some technology advancement or a new application, absolutely open to it. So just email me at yabdiche@carterra-bio.com.

 

Question

0:48:11.6 JB: Great, thank you Yasmina. Katie, a question for you. One of our audience members would like to know what other kinds of birds do you plan to make.

 

0:48:21.0 KC: Yeah, that's a great question. So we have actually several different lines of our heavy chain, as I mentioned, so the rearranging SynVH-SD bird, and also the pre-rearranged SynVH-C that I talked about. We have, in collaboration with quite a number of partners, are kappa bird, that we've also started to introduce our lambda bird, so carrying a human lambda genes. And we're also actually working on a common light chain bird, so a bird that's carrying a single light chain framework that has minimal mutation to it for... By specific antibody discovery. And actually, we chose the same framework as our OmniFlic, so there's a possibility there of combining antibodies from the rodent and also the chicken.

 

Question

0:49:11.3 JB: Alright, thank you Katie. Yasmina, we have another question for you. An audience member's asked for antibody screening at Carterra from crude lysate, how did you use the antibody as an analyte? Did you purify before the kinetic assay?

 

0:49:28.5 YA: So yeah, so for the kinetic, we don't use the antibody as the analyte, we use the antibody as the ligand. So we array it onto the biosensor using a capture method. So that is amenable to crude, you can take some hybridoma or lysates and then just capture them via a tag or via the Fc, and that's kind of like an inline purification. And then the analyte would actually be the antigen for those antibodies. The only example I can think of where you'd use the antibody as the analyte would be as an anti-ED. And in that case, the analyte would have to be a purified SAB, but the anti-ED themselves can be crude, they can be bivalent, and those would be the ligands.

 

Question

0:50:18.0 JB: Alright, Yasmina, stay with us. We have another question for you. One of our audience members would like to know, how is the level of immobilizations on individuals spots controlled in the 96-channel mode?

 

0:50:32.7 YA: That's a great question. So the 96 channels are in parallel, and so they are all the same in terms of the method used. So ways in which we would control the mobilization is to vary the concentration of the ligands, kind of titrate that, or we would titrate an activation level. So in the example where you have coupled ligands, you could lon one activation condition across the whole chip, or in individual channels, you could activate them with a dilution series and then present the ligands at the same concentration. So really concentration or potency of activation are two ways in which you can vary ligand density when you're looking at 96 in parallel experiment.

 

Question

0:51:37.0 JB: And we have another question for you as well, have you done experiment with 384 phage that display different FAB proteins and test analyte binding?

 

0:51:52.3 YA: So we do have examples of experiments that we've run from, say, periplasmic extracts. We wouldn't run the phage itself 'cause it wouldn't really be amenable to the SPR phenomenon, it would kinda just be too large. But we definitely run crude samples, again, the way in which we would do the array is capture out the crude sample via a tag, like Anti-V5 or a His-tag or FLAG-tag, some tag that is on the molecule. And as long as the analyte that is being bound is monovalent, then that would be a really nice assay.

Question

0:52:37.4 JB: And one last question for you, Yasmina, as well, how much of antibody 1 and antibody 2 sample volume would you need for a 96 by 96 epitope binning?

 

0:52:50.3 YA: Well, the brilliant thing about the array is that there is no scaling in the volume, so you would use the same amount of material to do a one-on-one as you would do a 384 on a 384, or a 96 on a 96. To create the spot itself, the ligand, we would use about 15 micro liters of antibody and we would couple usually at about two micrograms per mil, something like that. And then as an analyte, we would present again, 150 micro liters or so and usually we'd use about 5-10 micrograms per mil. So in total, I would say regardless of the size of the experiment, you could only use about two to five micrograms of your antibody in the role of both ligand and analyte. So it is really efficient on your antibody use.

 

0:53:47.3 JB: Alright, thank you Yasmina. And with that, we've come to the end of our webinar. So I'd like to remind everyone that the webinar will be archived on our site at www.genengnews.com for up to six months. So if you missed any part it, you can watch it again, or feel free to forward them your friends and colleagues, which we always recommend. I'd like to think Kathryn and Yasmina again, for their informative presentations, and I'd like to thank the audience for their attention and very thoughtful questions. And a very special thanks to Carterra responding this webinar. So hopefully, we'll see you again at another GEN webinar in the near future. Goodbye for now.