Scientists at Carterra, Twist Bioscience, ChemPartner and Berkeley Lights discuss modern-day high-throughput antibody discovery technologies that facilitate a paradigm shift in antibody screening and characterization—now discover therapeutic candidates in days vs. months.

Tom Yuan from Twist Bioscience presents on research published recently in Antibody Therapeutics. You will learn about the rapid exploration of the epitope coverage of a large panel of antibodies to guide the discovery of therapeutic antibody cocktails and combat infectious diseases. Tom also discusses integrating Twist’s synthetic biology technology with Carterra’s HT-SPR™ and epitope binning to quickly characterize antibody candidates.

Shireen Khan from ChemPartner presents on how the Berkeley Lights Beacon® Optofluidic System and Carterra LSA® facilitate the rapid discovery of therapeutic antibodies to fight cancer. Shireen discusses how plasma B cells secreting functional antibody candidates can be rapidly identified on the Beacon system compared to several months for a hybridoma campaign. Screening those antibodies for epitope specificity and affinity can be completed in just days on the Carterra LSA, thus substantially accelerating the antibody discovery process.

After the presentations, Anupam Singhal, Senior Product Manager from Berkeley Lights and Noah Ditto, Technical Product Manager from Carterra join Tom and Shireen for a Q&A session and panel discussion.

Presented by:
Tom Yuan, PhD, Senior Scientist, Twist Bioscience
Shireen Khan, PhD, Senior Director of Biologics, ChemPartner

0:00:00.5 John McKinley: Thank you all for attending today’s virtual event. My name is John McKinley and I will be your host. If you experience difficulty with audio or advancing slides, please refresh your browser window. Today, we have exciting presentations from Tom Yuan at Twist Bioscience and Shireen Khan from ChemPartner. First, Tom will discuss integrating synthetic biology technology with high throughput SPR to characterize antibody candidates in days versus months. Tom joined Twist Bioscience in 2018 as part of the biopharma vertical to drive antibody discovery and optimization by leveraging Twist’s unique DNA synthesis platform. He received his PhD from the University of California Irvine. After Tom’s presentation, Shireen will present on the rapid discovery of anti-PD-L1 antibodies from immunized Trianni mice enabled by B cell cloning on the Beacon platform and epitope binning with the Carterra LSA. After the presentations, Anupam Singhal, senior product manager from Berkeley Lights, and Noah Ditto, technical product manager from Carterra, will join our presenters for a Q&A session and panel discussion. Please submit your questions with the box on the right side of the screen. I’ll hand it over to you now, Tom.

Tom Yuan Senior Scientist in the biopharma vertical here at Twist Bioscience

0:01:17.6 Tom Yuan: Hi, my name is Tom Yuan, and I’m a senior scientist working in the biopharma vertical here at Twist Bioscience. It’s my pleasure to present today a case study where we used Twist’s synthetic DNA platform to rapidly synthesize, express, and purify hundreds of anti-Ebola virus antibodies. We then used the Carterra LSA system to probe the epitope landscape of the Ebola virus glycoprotein antigen. As you’ll see during this presentation, we find that by moving the epitope binning further up the antibody discovery workflow, we’re able to fully characterize a library’s epitope landscape rather than be limited to binning only a handful of leads.

0:02:02.5 TY: At Twist Bioscience, our core competency centers around our technology platform to print DNA onto silicon wafers. Using this platform, we’re able to synthesize oligos and oligo pools up to 300 base pairs in length in a massively parallel fashion at very low cost. Within Twist Biopharma, we use this capability, which we use to produce high-quality precise antibody libraries that not only give us control over the amino acid distribution in every position in the antibody, but also allows us to dictate exact sequences or motifs that we wish to include. Since we’re making each sequence explicitly, we can remove restriction sites, unwanted motifs, and sequences that may lead to developability issues downstream. This gives us vastly more control over traditional techniques to build diversity using techniques such as degenerate oligos or error-prone PCR. Additionally, we can easily incorporate multiple germline scaffolds and various CDR lengths in our antibody library designs. All of our libraries are validated by next generation sequencing to validate the final product, typically in the form of a phage display library, matches the initial experimental design.

0:03:22.0 TY: Shown here is the basic approach we use to build antibody libraries from oligo pools. The variable heavy and variable light domains of the antibody are comprised of three separate CDR loops spaced apart by constant framework regions. We synthesize each oligo pool to encode for diversity representing each of the three CDR loops. We then PCR the oligo pools together to create the antibody hyper-variable domains. Essentially, these are CDR shuffled libraries. When building antibody libraries, labs will typically use degenerate codons to build diversity. While this can mimic the antibody sequence distribution at specific positions in the antibody, it can still lead to unnatural combinations of amino acids that don’t match the natural CDR repertoire. Because we can dictate the exact sequences within these regions by printing the DNA, it allows us to design synthetic libraries that match the natural human repertoire and rationally sample from the desired sequence place. We can even include known binding motifs within these CDR regions. At the same time, DNA coding liabilities that may lead to isomerization, cleavage, deamidation or glycosylation sites are screened out so that they’re never present in the library to begin with.

Our work at Twist Biopharma

0:04:42.0 TY: As a general overview, our work at Twist Biopharma, which is again a division within Twist Bioscience, we’re focused upstream within the RD workflow and can assist our biotech and pharma partners to discover new antibodies or optimize existing antibody leads. We’ve produced and validated an entire suite of antibody phage display libraries that can either be licensed out to you or we can perform the discovery, expression, and the characterization work at our lab here in South San Francisco. We also have an antibody optimization platform that we call TAO, or which stands for Twist Antibody Optimization. Here we start with any antibody sequence as an input, whether human or not, and run it through our proprietary software that generates an optimization library that references a large human NGS database to improve affinity, potency, and developability.

0:05:38.3 TY: Here, I’ve listed most of the antibody libraries we’ve created here at Twist, which we like to call our library of libraries. These are comprised of naive synthetic libraries that can be used against any target, as well as target class-specific libraries, which can be used against difficult drug GPCR and ion channel targets. Among the naive synthetic libraries, the Hyperimmune Library is a fully human Fab library that utilizes the natural human diversity and incorporates over 2 million heavy chain CDR3s. There’s two versions of the Hyperimmune Library, the original version, which includes light chain diversity, and we’ve also constructed a Hyperimmune Library using a common light chain for use in common light chain bispecifics. We also have a suite of VHH single heavy chain libraries that are derived from published llama sequences to generate combinatorial diversity in both llama and humanized single domain frameworks. The Structural Antibody Library is derived from sequences from all published antibodies with well-defined crystal structures. The idea behind this library is that we’re sourcing from existing proven antibodies that are highly developable and stable by virtue of being amenable to generate x-ray crystal structures.

The process of developing and validating

0:07:00.7 TY: So in the process of developing and validating all these antibody libraries, we’ve also developed a high throughput antibody production workflow that we use in-house to express and purify hundreds or even thousands of antibodies, at both 1 mill and 8 mill scales. We’re starting to actually offer this now as an alpha product. And as our very first custom protein offering, anyone can now order the DNA encoding these antibodies but also order and receive the expressed and purified antibody itself that is ready to be used in any downstream experiment. We’re able to deliver both full-length IgG as well as single domain antibodies such as VHH-Fc. The antibodies are expressed in Expi293 and we typically see about a 150 microgram yield at the 1 mill scale and 700 microgram yields at the 8 mill scale. Typical turnaround time now is about 20-30 business days depending on the scale that is selected. Critically here, the 1 mill antibody expression product actually gives us enough reagent to perform high throughput surface plasmon resonance experiments to epitope in a panel of hundreds of antibodies. The delivery buffer is also an amine free buffer. And this allows us to directly use the purified product to amine couple these antibodies to the gold surface of the SPR chips.

0:08:30.7 TY: In this case study highlighting the use of high-throughput SPR to epitope in these antibodies, we’re building on the work of Zachary Bornholdt and colleagues at the Scripps Research Institute and at AdiMap, highlighted in the paper shown on the left. In their 2016 publication, they isolate and characterize potent neutralizing antibodies isolated from the survivor of the 2014 Zaire Ebola virus outbreak. So to fight infectious diseases, antibody therapeutics must achieve broad epidural coverage of the target antigen to enable potent, long lasting neutralization by complimentary mechanisms of action. Epitope binning organizes antibodies into epitope families by assessing the blocking relationships of pairwise competition to the target. In this case, the Ebola virus to glycoprotein. In traditional techniques such as FACS, and ELISA, or lower throughput label free biosensors, epitope binning using these techniques represent a few on many approach, which relies on existing antibody standards with benchmark antibodies to compare against.

Purified antibodies using our high-throughput IgG alpha product

0:09:40.2 TY: And in our paper, published this year is shown on the right where we synthesized, expressed and purified these same antibodies using our high-throughput IgG alpha product. We then applied them to a high throughput SPR-based epitope bin on the Carterra LSA, which allowed us to assign epitope bins in a many on many approach. Here, I’ve outlined the typical workflow and timeline to epitope bin any large panel of antibodies. To perform this workflow, all you need is the amino acid sequence of the variable heavy and light domains of the antibodies as well as a soluble protein target. In our case, we’re using the Ebola virus glycoprotein. Once we have the amino acid sequences in hand, we can back translate and reformat into HG1 for expression in our mammalian expression vectors. After a 14-day turnaround time, the clonal genes are then delivered as purified plasmid DNA product, ready for transient transfection in Expi293. We’ve developed a high throughput antibody expression workflow that takes about another four days for expression and purification.

0:10:52.7 TY: Once we have the purified antibodies in hand, then we can bring in the Carterra LSA to conduct two high-throughput SPR-based epitope bins. The first pass is designed to rapidly identify epitope bins in a non-optimized fashion where we don’t have to say normalized every single antibody concentration. Once the epitope bins are assigned from this first pass, representative antibodies are then selected as, “pathfinders” from each bin and then a second pass epitope bin is conducted using a larger scaled up antibody expressions to be used as higher quality reagents to recapitulate the entire epitope diversity of the library within a few key clones. So before I dive into the rest of the data, I’d like to take a step back and discuss why the Carterra LSA has played such a central role in the antibody discovery workflow here at Twist. So from the discovery stage to expression and purification and the characterization of all of our antibodies, all of our workflows here focus on high-throughput, highly parallel processes that the LSA really complements well.

0:12:08.3 TY: So during the course of a typical antibody discovery campaign here at Twist, we regularly reformat and test hundreds or even thousands of antibodies per project. And these antibodies are all reformatted as full-length IgG, or in some cases, the VHH-Fc construct. Once these antibodies are purified, we can use them and the Multi Full cell on the LSA allows us to print up to 384 of these antibodies to the chip surface. And it’s able to do this while accommodating sometimes lower antibody concentrations, so as low as 5 micrograms per mill. And it’s even nice enough to return most of the sample back to the plate after it’s done. So we’re able to reuse some of the remaining reagent. We utilize this throughput to characterize binding affinities as well as specificities and obviously, as well, is to perform epitope bins. But what’s really key here is that we’re able to perform these epitope bins as an early stage screen, instead of having to wait until later in the stage of the discovery workflow where we’re binning, say, only just a handful of antibodies. So for example, in a binning binding experiment, once the array of lean antibodies are mobilized onto the surface, which is shown here.

0:13:32.3 TY: On the slide up top, where you’re immobilizing, say, 96 or at the bottom, where you’re immobilizing 384. Once those antibodies immobilize each analyte antibody that we’re looking for blocking against, once that analyte antibodies float over the surface from, say, left to right here, it lets us capture 384 separate interactions at once. And really, it’s capturing more than 384, it’s also measuring against the reference channels, which is also key to the analysis downstream. So like I mentioned, other than the high throughput data collection, one of the main benefits of using LSA is its low analyte consumption during ligand coupling to the chip surface, the sample is actually cycled back and forth so that you’re able to set a coupling time of anywhere from, say, one minute or up to even 30 minutes or longer. But in both cases you’re actually only using a fixed 200 µL of physical sample to inject into the system. One can also utilize the parallel nature of the LSA to oversample and replicate sensorgrams, whether you’re doing a kinetics experiment or an epitope bin. And the Carterra is also, well Carterra the company has also built a company data analysis software that rapidly calculates and presents kinetic quantification and data binning in a fairly fast fashion.

High throughput SPR epitope binning workflow

0:15:07.0 TY: So back to the high throughput SPR epitope binning workflow, both the first-pass and second-pass epitope bins utilize a premix format, as the Ebola virus glycoprotein is a multivalent antigen comprised of a trimer of GP1 and GP2 heterodimers. The first-pass epitope bin is a rapid, non optimized setup where we don’t even attempt to optimize this concentration of the premix antibodies and we’re instead backlogging them in about one to 50. Depending on the size of the antigen and the affinities of your panel of antibodies, you need as little as about five micrograms of antibody per clone. And in this premix format, we target about a 4-fold molar excess of the antibody to the antigen in this premix here to make sure that we fully saturate all the binding sites.

0:16:11.9 TY: So here we have an example of a sensorgram plot showing the normalized binding responses of either antibody premixed with the Ebola virus glycoprotein or the glycoprotein alone. So in this case, the antibody 15984 is the ligand antibody and it’s immobilized to the SPR chip. The blue traces represent the binding response of the glycoprotein alone, which is consistent across many regeneration cycles. So when we take the same antibody, the 15984 antibody and premix it with the glycoprotein, we see that it fully self blocks. So this is just a good verification that the antibody is fully saturating the antigen solution and that we get a fully self block. So within this red binding region if with any binding response that ends in this red region, we denote that as a fully blocking interaction. So as you can see here, all the other antibody pairs, actually do not block binding to the Ebola virus glycoprotein. And actually many of them form prominent sandwiching pairs, which give us a response units well above just the antigen only.

0:17:38.0 TY: So if we continue this and generate… So in the full competition matrix, we’ve immobilized 233 unique antibody ligands to the chip and then injected 52 unique analyte antibodies premixed with the Ebola virus glycoprotein. So those 223 ligands are represented here in rows, so you see them each represented just as a single row. And then the 52 analytes are represented in columns here. And you can see that if you look at the blocking behavior of these analytes relative to these ligands, you see that they form these really neat and ordered epitope bins that are represented by these red portions of the heat map here. The reason why you’re seeing 52 analytes instead of, say, the full 233 analytes where you’d have a one to one relationship in terms of the competition matrix, is that we did have to remove a lot of the data points because like I mentioned, the first-pass epitope bin is non-optimized. So you we had to go through and come to comb through the data and make sure that you have full binding of all of your antibodies to not only the Ebola virus itself, but also that each antibody is able to fully self block with itself.

0:19:10.4 TY: So in many cases, we did have to remove some data points here. But even after cleaning through the data, we see the obvious emergence of seven distinct epitope bins. The heat map showing the competition matrix for these 52 uniques analytes and 233 unique ligands with the replicates removed. Each analyte and ligand pair, it is characterized as block, so red, or not blocked, green, and then there’s also the self, the intermediate, yellow, which were not fully determined fully blocked or not, and then self-blocking is dark red. It’s a little hard to see here, but in the blow up here, you can see that these dark cells here represent a full self-block of an antibody coupled with a chip, that same antibody bound to premix with the glycoprotein and demonstrating the rare implicate full cell, self-block.

0:20:18.8 TY: So if we zoom in on the lower portion of this heat map, looking at communities that we assigned as five and community 6 and community 7, we see evidence that these antibodies belonged to extremely rare epitope binning communities where only a handful of antibodies belong to the epitope bin or in some cases, there’s only a single antibody that represents that epitope bin. So, from here, from the first-pass epitope bin results, we select representative antibodies on each bin to serve as a pathfinder clone that are further investigated in a second-pass binning.

In second-pass binning, we combine our pathfinder antibodies with available benchmark

0:21:01.2 TY: So in the second-pass binning, we combine our pathfinder antibodies with available benchmark, anti-Ebola virus antibodies, in which we know where these antibodies actually bind to on the glycoprotein itself, and then we supplement that with our orthogonal data into a more focused binning experiment. So with these structural… I’m sorry, with these benchmark antibodies included, we can then infer the structural significance of any antibody assigned to a specific epitope binning from which those benchmark antibodies belong to. So for example, this ADI-15741 clone here, because it falls within community 2, we know that the community 2 antibodies include this map 100 benchmark and the KZ52 or FVM-09, which have already been described in a literature in terms of exactly where they bind on the glycoprotein, and, well, I’ll show that in a later slide as well.

0:22:17.1 TY: So, I also wanna make a note that, although we do make the use of benchmark antibodies here, anyone can run these epitope binning experiments with a completely new set of antibodies against a completely novel target, even in the absence of any known binders. We include them here just so that we can correlate and for more information from our epitope bins, but to generate the epitope… To actually discover the epitope bins themselves, you don’t need to know any structural information about your antigen. The epitope binning… The epitope communities actually emerge in the first-pass bin from the pairwise blocking relationships between the antibodies. So if you think about it, we’re kind of generating our own “benchmarks” through the selection of these pathfinder clones. Even if the pathfinder clones don’t intrinsically give you structural information on their own, they do at least tell you the… They really give you a key insight into the epitope landscape of your antibody libraries.

0:23:31.2 TY: So here, we’re comparing our epitope binning assignments by high-throughput SPR, those assigned by FACS in the Bornholdt et al paper. So the outer ring here in this Figure A represents the FACS binning assignments from the Bornholdt et al paper. You’ll notice that there are spaces of white that represent antibodies unable to be assigned epitope bins initially. The inner ring shows our epitope binning assignments from the high-throughput SPR experiments. And if you just look at this visually, you’ll see that the vast majority of the antibodies show agreement between the two binning techniques. In addition, we’re also able to assign some unknown antibody bins that were initially unknown and shown as white in the outer ring and we were able to assign them to antibody bins that we termed community 5 and 7.

0:24:35.3 TY: So I like this visual of ordering by the sequence dendrogram because you’re able to see visually that the distribution of bins along the sequence dendrogram is not… It doesn’t cluster perfectly. So that’s telling us that relying solely on sequence diversity is a really poor predictor of epitope. So this reinforces the fact or it reinforces the use of epitope binning as a screening tool early in the antibody discovery workflow. So just because you have really… If you have really high sequence diversity, you have two clones that are, say, really, really completely unrelated in their sequences, they may still bind to the exact same epitope bin.

0:25:30.1 TY: So if we continue along, figure B shows the antibody distribution in terms of how many antibodies belong to which communities, and we see that communities 2 and 3 represent the most immunodominant epitopes. And then figure C shows us the break down of the antibodies among the different bins and how they correlate to live virus neutralization. So, for example, in community 3, many of these antibodies gave us very potent antibody neutralization. Sorry, these antibodies confer very potent live virus neutralization as opposed to bins 5, 6 and 7, in which, say, half or less than half of the antibodies give a strong neutralization.

Additional benefit of using SPR to epitope bin

0:26:17.9 TY: So an additional benefit of using SPR to epitope bin is the wealth of data that you collect in the process of generating these blocking heat maps that we’re showing earlier. So, in this slide, we have identified… We’re showing Sensograms where we can identify several nuanced blocking behaviors just by looking at these Sensograms.

0:26:40.3 TY: So, in figure A, if we look at the Sensograms where 15841 is immobilized in the chip, 15841 fully self-blocks, when you look at binding of the glycoprotein alone, you see consistent binding of… Binding behavior of the glycoprotein alone, and then you see that mAb114 does not block and it belongs to a separate community. But if you look at the blocking behavior of 15878, it does appear to fully block with 15841, which belongs to community 1. But with KZ52, which belongs with community 2, when you mobilize kZ52 out of the chip, you see that this ADI, the same antibody 15878 seems to block as well with kZ52, so 15878 blocks across community 2, with kZ52 and also community 1, shown here by 15841. And just to double-check that with mAb114, mAb114 does not block either community 1 or 2, and as such, it shows this very large sandwiching behavior across these two Sensograms.

0:28:01.7 TY: In Figure B, we see evidence of order-dependent blocking asymmetry of FVM-09 with community 2 members. So, with F… So, in this case, blocking is only observed when FVM-09 is allowed to bind to Ebola virus first in the premix step, and this actually agrees with the literature reporting that FVM-09 acts cooperatively in antibody cocktails due to a triggering of an induced epitope effect, where binding FVM-09 will expose… Well, upon binding with glycoprotein will expose additional binding epitopes. So, in this case, we see that, with FVM-09, when ADI-15741 is allowed to mix first with the Ebola virus glycoprotein, it does not show any blocking behavior to FVM-09 coupled with the chip. But we reversed this assay, and we allow FVM-09 to mix with glycoprotein first in the premix, we do see that it actually does block when 15741 is immobilized in the chip, so this is showing an order-dependent blocking asymmetry.

0:29:35.3 TY: In figure C… Figure C hits at possible antigen heterogeneity of the glycoprotein itself actually, so that may hint at two possibly distinct structural populations of the glycoprotein and this is evident because we see, if we look at the blocking behavior of where 15779 is ligand is immobilized in the chip, so 15779 belongs to community 4. So it fully self-blocks the glycoprotein binding responses as expected, but when you’re looking at the blocking behavior with KZ52 premixed with the glycoprotein, it behaves exactly like the glycoprotein alone, and when you reverse this assay and you immobilize KZ52, and instead premix a flow over the premix of 15779 with the glycoprotein, once again, it looks exactly like… The binding behavior is exactly like EboV glycoprotein alone. And what this hints at is that basically there are two separate populations, or there might be two separate populations of the glycoprotein in solutions such that these pairings are completely… Do not affect binding at all with each other at all, in either format.

In figure D, we see that rare epitope assignments

0:31:05.5 TY: And then in figure D, we see that rare epitope assignments are really bolstered by this high throughput SPR technique by seeing that… We’re seeing repeated… Repeating at the assay over over again to confirm that there’s only self-block of 15983 with itself, and there’s no other antibody in this set that is able to block that interaction. You see this repeated over and over again with the vast number of traces, Sensograms traces that you see that are either equivalent to the Ebola virus glycoprotein alone or sandwiched, hence give you greater normalized binding response.

0:31:58.0 TY: So here we have the crystal structure as shown of the prefusion trimeric Ebola virus glycoprotein without the mucin-like domains and the transmembrane domains. So the structural epitopes are color coded, and then the benchmark antibodies along with the communities have been assigned and highlighted for their various different epitopes. So if we map the epitope footprint of the structural benchmarks, we see that the epitope communities belong to large exposed epitopes, generate large numbers of community members such as community 3, while smaller number of antibodies bind to the more occluded epitopes, such as the GP1Core covered by community 5. So, if you remember, from earlier, there are many, many antibodies that fell into community 3, but very, very few that fell into community 5, and we see those represented in the crystal structure itself. So, in general, fully-characterizing the epitope landscape of the antibody library is critical to developing neutralizing antibody cocktails that may leverage cooperative mechanisms to help neutralize the viral agent and it may also minimize the risk of meningitic escape as well.

0:33:25.9 TY: So with that, I’d like to thank you for your time, as well as our collaborators at Integrated BioTherapeuics who provided the glycoprotein and various control antibodies. I’d like to thank Carterra as well for inviting us to speak today, and I’d love to take any questions you might have during the Q&A session.

0:33:46.7 JM: Thank you, Tom. Please make your questions for Tom with the box on the right side of the screen. We will answer the questions in the panel discussion after our next presentation. Now, I would like to introduce Shireen Khan. Shireen is a Senior Director of Biologics at ChemPartners, South San Francisco, where she leads a group that has expanded ChemPartners’ capabilities into single B cell cloning on the Beacon platform. She also leads multiple therapeutic antibody discovery programs for biotech and pharmaceutical companies. She completed her PhD in Biology at the University of California, San Diego. I’ll hand it over to you now, Shireen.

0:34:26.8 Shireen Khan: Thank you, John, for that introduction, as well as the opportunity to present this story today, and thank you to the Merck folks for listening in on this talk. I’m going to present a story, a collaboration project, really, between Trianni and ChemPartner, and we also collaborated with Carterra to generate some of the data I’m gonna present today, where we were really trying to bring together different technologies to evaluate the accelerated discovery of anti-PD-L1 antibodies, so we used the Trianni mice as well as the B cell cloning capability that’s now at ChemPartner on the Beacon platform, and we use the Carterra LSA to be able to characterize our antibodies further for affinity as well as epitope binning.

0:35:24.6 SK: So, the value and impact of drug development has really come to the forefront with the current COVID-19 pandemic situation, where it’s really becoming obvious that not only do we need to accelerate the timelines needed to generate effective therapeutic strategies, but that we should really, as an industry, kind of engage in more collaborative efforts, as we’ve seen in the past several months. We’ve been able to see resources, funds, as well as expertise, coming together between large pharmas, small biotechs, universities, as well as CROs, towards this effort to really bring high quality therapeutics to those patients who need it most.

0:36:13.4 SK: ChemPartner is actually collaborating with Gladstone and UCSF to discover drugs against SARS-COV2 by multiple modalities. I won’t present any of that work today, but rather, really just thinking about this process of drug discovery and there’s an experiment unfolding in front of all of us to see how long does it really take to discover, develop, test and approve an efficacious drug? And so, there’s really a need for us to take a lead in figuring out, how can we drive and promote innovation and leverage new technologies? In what way can we accelerate the process and deliver these much needed therapeutics to the patients?

0:36:58.4 SK: And so, in this talk, I mentioned in the beginning, we’re gonna really talk about how we were able to leverage the Berkeley Lights Beacon platform, transgenic animals, specifically the Trianni mice, as well as the Carterra LSA to kind of push forward the discovery of, in this case, anti-PD-L1 antibodies.

ChemPartner is quite invested in accelerating the antibody discovery process

0:37:32.8 SK: So ChemPartner is quite invested in accelerating the antibody discovery process. So, our normal hybridoma campaign could take up to 20 months, going all the way from reagent generation through assay development, the immunization, the hybridoma development, sequencing, humanization, and then finally the manufacturability and large scale production. That whole process takes about 20 months, so we wanted to invest in discovery capability that could reduce that cycle time, and so with B cell cloning using wild type mice, for example, we were able to reduce that amount of time by up to six months. So that was a significant improvement right there, but then we took it further to ask the question, okay, if we use a transgenic animal, in this case, the Trianni mice, we could reduce the cycle time even further down to 10 months because we’d be able to forgo the humanization process.

0:38:44.0 SK: So just wanted to introduce the Trianni Mouse platform. As you probably are very aware of, the mouse has the entire human antibody repertoire contained, so it has the human Ig gene segments that were precisely targeted in place of the deleted mouse segments at all three loci, the heavy, as well as the kappa and lambda light chains. So there’s unique DNA engineering that really combines biological function with FTO, and these mice are immune competent, they have normal B and T cell development as well as immune responses.

0:39:30.1 SK: So the structure of the gene there is shown on the left as an example, you can see the human VH gene segments in the normal human loci, and so what happens here is that the human VH gene segments in this case, are actually swapped in to the mouse non-coding DNA. And so what you end up getting is the native mouse regulatory elements driving human antibody repertoire production, and what you end up with is the VDJ recombination, as well as somatic hyper-mutation and heavy chain class switching, which would be comparable to a wild-type mouse. And the Trianni Mouse, in particular, has comparable properties to those obtained from a normal wild-type mouse with respect to sequence diversity, binding affinity, epitope coverage as well as specificity, and there’s a low risk of immunogenicity as well as what they’ve seen is good manufacturability properties.

0:40:39.0 SK: So on the Beacon platform at ChemPartner, we have been able to utilize this platform across multiple campaigns, and we are able to load single B-cells onto these OptoSelect chips, which are shown here in the middle, which basically allows us to introduce and load single B-cells into these nano pens, allow them to culture and secrete antibody, which allows us to assay for target-specific binding as well as functional activity within these nano pens, and so, within a single day, we’re able to identify target-specific binding and function of tens of thousands of single B-cells, and it really essentially is moving function forward in the drug discovery process, which we believe is very important for accelerating the drug discovery process.

The collaboration projects between ChemPartner and Trianni

0:41:47.1 SK: So the collaboration projects between ChemPartner and Trianni, we basically kept building upon the previous collaboration project that we had with Berkeley Lights, where initially we compared the output of a hybridoma campaign to that of B-cell cloning using a wild-type BALB/c mouse. And so that was our first anti PD-L1 case study, and in this study we aim… Our aim was to basically accelerate that further using the Trianni Mouse platform using the Beacon single B-cell cloning capability, and we wanted to look at two different immunization strategies, again, with that theme of trying to reduce cycle times, we wanted to look at the output from Hock immunization strategy compared to IP immunization, so the Hock taking about 31 days total, compared to IP, which could be anywhere from 60 to 74 days.

0:42:48.9 SK: And so, after each of these immunization campaigns were completed, we isolated the plasma B-cells using magnetic CD138 isolation kits, and then loaded our B-cells onto the Beacon and then within that same day, screened for anti-PD-L1 specific antibodies that blocked the binding of PD-1 to PD-L1. So we basically were able to use a variety of the functional capabilities available to us, either in our ChemPartner Shanghai facility or in ChemPartner South San Francisco location. So in our Shanghai facility, we’re able to leverage the Protein Sciences group there, which was able to generate the recombinant proteins as well as generate the cell lines and the recombinant antibodies. We also had all of these assays developed already in Shanghai and just transferred them and translated them onto the Beacon platform.

0:43:58.3 SK: And of course, the B-cell cloning was located in our South San Francisco location, but once we did the sequencing, we were able to do manufacturability assessment as well as functional assays back in Shanghai, and of course, the epitope binning was actually done out of the Carterra lab in Dublin. So the Beacon screening strategy for this campaign was actually to do not only cell-based, or we actually did B-based binding in the channel as shown here, but then we also did cell-based binding as well as blocking assays within the pen. So the versatility on the Beacon platform allows us to be able to screen for target-specific binding in the channel, and that’s what’s shown here, so we have our plasma B-cell loaded along with jurkat cells over-expressing human PD-L1 and then we would flow in beads that were coded with human PD-L1, and by adding a detection antibody, we basically saw a bloom above the pen where that particular plasma B-cell is secreting a target-specific antibody.

0:45:22.4 SK: And so in order to evaluate the binding and blocking activity, we basically had a very similar set of reagents, except we used an anti-mouse IgG detection antibody, Alexa Fluor 488 labeled, and then we used a soluble PD-1 that was PE labeled. And so what we are basically looking for is whether there would be Alexa Fluor 488 signal on the cells, which is actually what’s shown here, so this definitely indicates there’s a binding antibody present, and we were looking for whether or not the human PD-1 would accumulate at these cells, which you can see very clearly here, suggesting this is a non-blocking antibody, but there’s no accumulation in this example here suggesting this is a binding blocking antibody, and that’s how we selected our hits for the…

We screened over 25,000 B-cells and identified 135 antibodies

0:46:23.0 SK: So, at the end of this campaign, we screened over 25,000 B-cells and identified 135 antibodies that block the binding of PD1 to PDL1. And of these blocking antibodies, 93 of them actually came from the hock immunization strategy, suggesting that we could accelerate the discovery process by leveraging this immunization strategy that took a fraction of the time. We did a quick sampling, so we sequenced a total of 22 of the antibodies and recovered good VH/VL sequences from 20 of them, and all 20 of them were unique. We did a hotspot analysis, and 18 of the antibodies were selected for expression and purification. And the purified IgGs, this was all done in Shanghai, were characterized for cell-based binding as well as blocking. We did affinity by octet, I won’t show you that data today, but we did also affinity and epitope binning by the Carterra LSA, and I’ll show you that data today, and this was all compared to benchmark antibodies as well as a select set of antibodies from the wild-type BALB/c single B cell cloning campaign as well as that hybridoma campaign that we had done side by side.

0:47:57.5 SK: So this is a summary of the antibodies that we were able to obtain, so 20 out of the 22 had unique sequences, and there was normal class switching, and the kappa-lambda light chain ratio came out as expected with the majority of the antibodies being kappa. And so you can see for this clustering tree at the bottom here that a majority of the antibodies were unique and distinct from the benchmark antibodies as shown in this clustering tree.

0:48:40.0 SK: So, in this slide, we’re looking at the binding of the purified antibodies now, by FACS, and so this is looking at the binding to CHOKI1 overexpressing human PD-L1, and compared to on the right here, we’re looking at the benchmark antibodies as well as some of the Trianni antibodies, and you can see that the majority of these antibodies have very good binding profiles and they’re very similar to the benchmark antibodies.

0:49:16.7 SK: In this slide here, we are actually looking at the FACS binding profiles of cells overexpressing Cyno PD-L1, and you can see similar to the human, binding profiles that a lot of the Trianni antibodies showed really good binding, in fact to Cyno PD-L1 as well. So, we did not screen for Cyno cross-reactive binding on the Beacon platform, it’s something we could have done, but what we are seeing here is at the end of the screen, we actually have a lot of them with really good binding profiles.

0:49:57.5 SK: So, these are the results for the protein-based receptor ligand-blocking assay, and on the right, what we’re showing are the benchmark antibodies, which is basically showing that the benchmark antibodies, the BMS as well as Atezolizumab and Durvalumab, they all have a very good blocking activity and our best antibodies from the wild-type B cell cloning as well as the hybridoma campaign are right there in the middle, right in the sweet spot of these benchmark antibodies. And you can see similarly, if you look at the IC50 values here of the benchmarks, you can see that a majority of the Trianni antibodies shown here are within the range of the IC50 values of the benchmark antibodies, suggesting that we were able to pull out some very good potent antibodies from this campaign.

0:51:02.1 SK: So, our best two antibodies are shown here, 92 as well as 87 in blue. And I wanted to call your attention to mAb95 in orange here, you can see the IC50 value slightly right shifted compared to 87 and 92. And, we get asked this question quite a bit, so we went back and looked at our screening data on the Beacon, just to ask the question, “Would it have been possible to tell the difference between a clone such as 87 or 92 and that of 95?” So you’ll see that in the next slide.

0:51:44.3 SK: So, this is a retrospective analysis of the mAb95 blocking data on the Beacon. And unfortunately, you cannot really see a difference between a clone such as 95 and then the others that were even more potent, you basically don’t see a signal here. So, it kinda suggests that it’s an all-or-nothing signal on the Beacon. We’ve not gone back to look at the MFI values very carefully, to see if there was a slight hint of a difference, but, just to kind of be aware that really, when we run these assays on the Beacon, it’s really kind of a high content imager, it is gonna return an image and then you kind of have to make your calls about what you’ll consider a hit in real time by looking at these images.

Collaborating with Carterra to be able to do some epitope binning

0:52:42.4 SK: Okay, so at this point we were collaborating with Carterra to be able to do some epitope binning, as well as affinity estimations. So this is the LSA, which clearly is a way that folks can use in order to support antibody characterization, and it really enables high level of screening using very minimal sample volumes. And so the name of the LSA really alludes to its purpose in helping scientists to basically navigate large pools of candidates and identify optimal and nuanced properties. So it comes from the name loadstar, which is a reference point used for navigation, and the A is for array. So the way the system is configured, it’s really unique among biosensors because of its ability to switch the fluidics modes between arraying of up to 384 ligands followed by single injection across this array in a single channel mode. And from each injection, you get real-time binding signals measured for these 384 reaction spots plus 48 separate inter-spot references, and so the bi-directional flow that’s used in creating the array and the associating analytes during binding analysis allows for really robust signals to be measured out of very, very low concentration samples, such as what you would find in accrued supernatant such as hybridoma.

0:54:34.6 SK: And so, there’s versatility on this platform in terms of the different applications, so you can actually do affinity measurements, either by looking at global kinetics or steady state affinity. You can also do epitope characterization, which includes epitope binning, as well as epitope mapping, which can identify domain or even residue-level binding sites. And lastly, you can do determination of antibody concentrations, such as what you’d find in a hybridoma supernatant, and that can be quickly done using the quant workflow.

0:55:17.6 SK: So what we did with Carterra on the LSA in this case, was to do epitope binning of all of the Trianni antibodies, so it was 19 total, and then we included seven antibodies from our wild-type B cell cloning campaign and nine from our hybridoma campaign, as well as the three benchmark antibodies. And so there was a covalent array prepared using three different antibody concentrations, and then that array was first used for binding kinetics using five concentrations of the PD-L1 ECD. And then the same array was used to run basically the pairwise sandwich epitope binning experiment. And this experiment ran for 24 hours, which generated over 9000 sensorgrams, and it would have taken a much, much longer to do this using a Biacore.

0:56:23.9 SK: Okay, so we had this iso-affinity plot in order to really kind of evaluate what we ended up with for the Trianni antibodies compared to previous campaigns, as well as the benchmarks. And the sensorgrams are shown on the right here, which shows that this is the benchmark antibody, atezolizumab. And you can see we had two Trianni antibodies that looked very similar actually to this benchmark antibody, very fast on rate, very slow dissociation curves, and we actually had several antibodies from this very brief campaign that had single-digit nanomolar affinity or better, so it was quite a good output from a single campaign.

0:57:17.0 SK: So this is an evaluation that was done by Dan Bedinger at Carterra, just looking at the network plots. And on the left-hand side here, the epitope binning results are shown and they’re colored based on the antibody source, so apologies for all of the different colors going on in this graph here, but in red are the Trianni B cell cloning antibodies, and that represented six different bins, whereas when we first did this by B cell cloning using wild type animals, we mainly were clustered into four bins. And then the hybridoma campaign antibodies, actually in that initial study was in 13 different bins, but quite a lot of them had no activity in terms of the potency, and so we really just kind of selected the bins here that had the most potent blocking activity. And we were happy to find that a lot of the Trianni B cell cloning antibodies also had very high, or actually very good potency in the blocking assay here, so you can see this is kind of colored based on potency. And so a lot of the antibodies that are red over here kind of fit within the range of our benchmark antibodies, which extend anywhere between shades of yellow all the way up to red. And so a lot of these red antibodies here kinda fit into that category of having fairly good potency in our blocking assay. And so this analysis really kind of gave us a nice little glance at where all of these antibodies binned with respect to the different strains of mice and the different campaigns.

0:59:18.1 SK: So in this slide, I thought it would be really kind of interesting to just kind of highlight, looking back at the clustering tree based on sequence diversity, it was interesting that there are a few communities called out by the LSA that kind of starts to hint at sequence similarities. So community nine here as well as six and four clustered together, which was similar to what we found from the clustering tree based on sequence, which is really a very interesting point. It didn’t always correlate. Obviously, community two here, there’s quite a lot that were distinct in terms of sequence diversity, but it gives you some sort of hint at potentially, the diversity of the candidates. And when you’re doing the screening early, such as with hybridoma supernatant, that information can actually be very useful earlier in the process, when you don’t have the sequence information. Luckily, on the Beacon we’re able to get sequence and lineage information earlier in the process, which we think, again, lends itself to accelerating the discovery process and really focusing in on the best candidates earlier.

1:00:42.9 SK: So this is just a summary of all of the data, looking at the affinity ranking of the Trianni antibodies, and you can see highlighted with the blue arrows is the benchmark antibodies as well as all of the Trianni antibodies. A majority of them fit within six epitope bins, and they’re all represented in this range of single-digit nanomolar affinity or better. So we’re actually getting high-affinity antibodies stretching across multiple epitope bins within this very small set of antibodies that we pulled out to evaluate further.

1:01:31.3 SK: So in this slide, we’re actually looking at the antibodies with respect to the blocking assay. So these are ranked on IC50 values in the blocking assay, and this time I’m highlighting where the benchmark antibodies are, so the really potent ones shown here, as well as the BMS antibody shown here, and a majority of the Trianni antibodies actually fit within this range. So 14 of them were as potent or even more potent than the benchmark antibodies, and then two of the best antibodies, 92 and 87 were as potent as atezolizumab. And so if we had actually gone back and done deeper screening of the 135 blocking antibodies that we did initially identify on the Beacon, it’s likely that we could have pulled out even more antibodies with possibly greater blocking activity, or even affinity.

1:02:32.2 SK: So in summary, we screen nearly 25,000 single B cells isolated from the immunized Trianni mice. We had the greatest number of blocking antibodies identified from hock immunization, which is obviously faster than an IP strategy. A total of 135 blocking antibodies identified, and we were able to recover good quality CDNA from about 89% of the single B cell hits. Again, this was a quick sampling. So 20 out of the 22 antibodies were unique, and 18 were selected for expression and purification based on limited liabilities after the hotspot analysis, which in having that sequence information allows us to focus in on.


1:03:20.2 SK: We had 16 antibodies that bound to Human as well as Cyno PD-L1 over expressed on CHOK1, and then 14 out of the 16 had blocking activity comparable to the benchmark antibodies. The antibodies represented six distinctive epitopes, which is actually more bins than what we observed with the wild-type BALB/C mice, and then 10 of those antibodies had single-digit nanomolar or better affinity based on Octet as well as the LSA data.

1:03:55.8 SK: I just wanted to say, I’ll leave you with some final thoughts here. Definitely, there is a need to accelerate cycle times. And in this talk, we’ve touched on a couple of different strategies. So either the immunization strategy being hock saved us about a month, and using transgenic animals, of course, also can cut down on the time needed to otherwise undergo humanization. The B cell cloning on Beacon can save three to six months, depending, again, on the immunization strategy, and moving function forward in the drug discovery process enables us to focus in on the candidates that have the desired properties as early in the process as possible. We’re able to do sequence as well as epitope binning on the LSA and get that lineage information earlier in the process, which really, again, can help in identifying diverse, unique candidates early on. We can incorporate the hotspot analysis, which can also save time down the road to avoid antibodies with liabilities.

1:05:08.1 SK: And so it really kind of is begging the question, what other technologies can we leverage to further accelerate the process? I’m sure there’s more technologies out there that we could think about, but in the past, when going from basic, to translational, to pre-clinical, and then to clinical development, it’s been a very long cycle time. And so I think one of the questions we have to ask ourselves is, what will be the new normal for the drug discovery cycle time? It probably can’t continue to be as long as it has been because we’ve now very recently been able to see how quickly we can generate therapeutic antibodies against the SARS-CoV-2. So progress has definitely been made. So I wanna thank… today, and just acknowledging the folks in Chem Partners, South San Francisco for their contributions, as well as Shanghai, the Trianni team, Bao Duong as well as Linda Masat. And the efforts at Carterra, Dan Bedinger, Christian Giddens, as well as Tim Germann for all their support in this collaborative effort. So thank you very much for your attention and looking forward to answering any questions.

1:06:28.5 JM: Thank you, Shireen. Thank you, Tom, for your excellent presentations, there’s a lot of great information shared there today. Now we welcome our Merck team join in on a Q&A panel discussion with Tom, Shireen, and now joining us is Noah Ditto, our technical product manager from Carterra, and also Anupam Singhal, the Senior Product Manager at Berkeley Lights. Please submit your questions on the box on the right side of your screen. I see we have some questions coming in. Tom, can you describe how high throughput SPR compares to FACS analysis for epitope binning, how the assays scale and fit into a discovery workflow?

1:07:29.4 TY: Right, so for comparing high throughput SPR on for epitope binning versus FACS, I think the biggest takeaway there is that with the high throughput SPR bin you can just by the sheer number of spots that you can analyze on the chip, you can essentially run a pairwise competition matrix with the SPR. With the FACS analysis, typically that’s orders of a magnitude slower, you’re typically doing a many on one assay versus a complete pairwise matrix, so it also forces you to run the epitope binning if you run it by FACS, later down the discovery workflow and among a smaller set of antibodies as well. So if you can move it up further up the workflow use can use it more of a screen, it’ll let you capture much more of the epitope landscape.

1:08:29.1 JM: Thank you, Tom. There’s a follow-up question to that. Were you surprised to see antibodies from the same germline families binning to different communities?

1:08:41.3 TY: I don’t… Well in the terms of the Ebola antibodies, no, because we’re obviously using the established set of antibodies that are already discovered, but in other types of infections such as say Staph aureus or yellow fever virus, that in those cases, we see that the initial relation is much more aligned to specific germlines, much more germline encoded. In the case of Ebola, we don’t see that for this pathogen. So I think the takeaway here is really that you do have to test and you do have to test this and actually see what the data show you. It depends on… It really depends on the pathogen, from pathogen to pathogen.

1:09:38.5 JM: Oh, I see. I have a question here for Shireen. What are the most challenging aspects of functional assay development on the Beacon platform?

1:09:51.6 SK: Yes, great question. So one of the most challenging aspects is actually how robust the assay is, so if there’s already an assay developed by ELISA or FACS and you have a pretty good assay range and signal window, then typically that tends to be very representative of what you can get on to the Beacon. And so what we do is basically QC everything by FACS and then evaluate that and translate it back onto the platform, so really the key elements are the quality of the reagents going into that assay and QC-ing them ahead of time, and then being able to put it on the platform so that you can actually see if the signal is strong enough to be able to call your hits.

1:10:46.0 JM: Thank you Shireen. Do you have a sense of for the frequency of false-negative blocking antibodies on the Beacon?

1:10:56.0 SK: Yeah, so that’s a really important question as well, so for our initial wild-type BALB/C campaign on the Beacon platform, we actually went back and pulled… So we got a total of 35 blocking antibodies in that campaign after looking at 33,000 single B cells. We went back and pulled another 30 that were negative on the Beacon, and so we went back and express and purify, which actually was a lot of work, but we wanted to get to that question actually, and found that a majority of the ones that were negative were actually not great antibodies, some of them did have some blocking activity, but it was far weaker than the ones that we called out as hits in that initial screen. So we don’t feel that the false-negative rate is very high.

1:11:54.0 JM: Oh, I see. We have a question here for Noah. Are there technical differences in screening affinity for hybridoma versus B cell-derived antibodies?

1:12:08.4 Noah Ditto: Yeah, so the source of the antibodies, particularly, in this case, coming from a more accrued matrices really doesn’t matter in terms of setting up the experiment on the LSA. So on the LSA, typically, we have our antibodies down on the surface for kinetic screening type assays. And in this scenario, when we put them on the surface in a capture format where we’re flowing the supernatant across the surface to enrich, so effectively all the background matrices that the maps are encompassed in, all gets washed away before the antigen is introduced. So ultimately, we do on chip purification prior to our kinetics series, resulting in kinetics that are perturbed by the source of the antibody such as accrued matrices for example.

1:13:03.5 JM: Thank you, Noah. Here’s a question for Anupam. Do you have any other examples where the ligand receptor blocking assay on the Beacon has been used to down select need candidates?

1:13:18.8 Anupam Singhal: Yeah, absolutely. So it has been used in multiple other instances, probably the most notable recently was in workflows run at Vanderbilt University, where convalescent human patient samples were used to find blocking antibodies for SARS-COVID-2 infection. Those antibodies have since advanced into phase three trials at AstraZeneca.

1:13:49.1 JM: Nice, that’s great. Here’s a question for Tom. Can you talk about the mechanism of action of antibodies and how that determines their ability to neutralize viruses?

1:14:08.1 TY: Right, so for this set, the mechanism can be as simple as blocking the specific sites on which the virus interacts with the cell surface receptors, and in some cases, we also see instances where actually a combination of two separate antibodies that on their own are not neutralizing can become synergistic and actually enable neutralization. In other cases we see, such as with the SDM 09 variants or clones that binding of an antibody will induce a confirmational change, which will expose additional epitopes as well. And it’s also important to note that the… Being able to understand the whole epitope, the epitope coverage as much as possible is important to help in… If you want to introduce a therapeutic as a combination, so for example, this would help mitigate viral mutagenic escape. And you see this as well in some of the SARS-COVID-2 therapies as well. For example, the Eli Lilly therapeutic is a combination of two separate antibodies.

1:15:30.9 JM: Thank you, Tom. A question for Shireen. Was there a correlation between epitope bins as predicted by the LSA network plot and antibody sequence diversity?

1:15:49.5 SK: Yeah, so it’s interesting, and I’d like to hear Tom’s thoughts on this as well, but for this Trianni campaign as well as the wild-type B cell cloning campaign, there were some… There was a bit of overlap between epitope bins and sequence similarities, which I was highlighting just briefly in one of the slides there, but it didn’t always correlate. So there was that one community too in this slide deck that I presented, that they all fit into that community together, but they’re very diverse candidates, so I think it can correlate at times, and then other times, maybe not so much, but I think it’s a really interesting property. And so maybe Tom has more to add on that.

1:16:49.6 TY: Yeah, so that’s a very good observation. And now what we’re seeing is that it really depends on the target. In terms of the infectious disease space, we see that more common infections or pathogens that are more associated with opportunistic infections, such as staph aureus, so commensal pathogens, they tend to be more strongly… The neutralization tends to be more strongly germline-based, but we don’t see that in Ebola. And also in our hands as well, depending on the different targets for antibody therapeutics, not necessarily for infectious diseases, we do definitely see different behaviors between the similarity of the sequence and the epitope bins themselves as well. I think at the end of the day, for us the take home message is that the more data we have initially with epitope binning, the more we’re less reliant on, solely relying on say sequence data or the lead data or even in vitro neutralization data to drive home the lead selection because in some cases, yes, relying on sequence diversity may help, may correlate with epitope diversity, but many other cases, you don’t know, and unless there’s already literature out there suggesting one way or the other, you’re kind of flying blind, which you obviously don’t want… You don’t wanna be doing.

1:18:29.9 SK: So maybe, Tom, one follow-up question to that is, so the differences that you did see between the different infectious diseases, do you think it has to do as well with the target and just how much is exposed and available? Because for certain GPCRs or ion channels, when you have just very limited ECD, maybe there is more chance that there’s… You’re gonna pull out the same type of antibodies because you just don’t have as much real estate there. So what are your thoughts about that?


1:19:06.7 TY: Yeah, that’s definitely a huge factor as well, in terms of how much actual surface area is exposed and available to the antibodies as well, that’s a completely valid point. I think in the case of the Ebola virus glycoprotein itself, it’s quite a large target, so in this case we’re seeing there’s much, much more epitope diversity in terms of the antibodies we’re able to pull out. We have seen in other cases where the targets are much smaller, where everything will collapse into one or two epitope bins, so you’re correct, it’s highly dependent on the targeting and the expression of the targeting on the cell surface.

1:19:56.9 SK: Right.

1:20:01.7 JM: Thank you, Tom and Shireen. Anupam, what new features exist in the antibody describing workflow on the Beacon?

1:20:14.6 AS: Yeah, one of the new features that we released this year is integrated on-chip genomics capability, we call them the GPCR, which enables on-chip CDNA synthesis to improve the efficiency and throughput of sequence recovery on the system.

1:20:38.6 JM: Thank you, Anupam. There’s a question for Noah. Can you perform epitope binning in kinetics using the same chip?

1:20:52.1 ND: Yeah, that’s a good question. Potentially, yes, it kinda comes down to what your objectives are and sort of what you’re trying to accomplish with the data. So in kinetic experiments, there’s typically an interest in keeping the density of the species on the surface called a ligand, typically antibody on the surface, fairly low to facilitate really good kinetic properties, ligand affinity and sort of other non-ideal kind of interactions, while in binning generally, the density of the surface is not as much of a concern, it’s more of an effort just to have a robust signal, typically over the course of many cycles, so you can put them together into a single assay and definitely get double duty out of a single surface. It may come that you sort of strike a balance between the two assays. You may find that having a little more surface density in your kinetics assay to support the binning is probably advantageous and certainly it’s an exercise of trying to relatively screen and rank clones prior to binning. Certainly, that’s a suitable kinetic endeavor to set up the assay in such a way that you have appropriate surface densities for the binning and still get meaningful kinetics prior to that.

1:22:17.7 JM: Thank you, Noah. Well, that was our last question. And we’ve reached the end of our time. If you all have any additional questions, please reach out to Tom, Shireen, Noah and Anupam. Their email addresses are listed on the screen. I would like to thank our presenters, Tom and Shireen, our panelists, Noah and Anupam, thank you all, and I hope you have a fantastic day.