In this webinar, Derek Croote, PhD from IgGenix and Lucy Liu, PhD from Alloy Therapeutics discuss new advances in HT-SPR, single-cell BCR-seq, and RNA-seq.

Although it is highly rewarding to create medicines, particularly therapeutic antibodies, doing so comes with several risks. It is expensive, time-consuming, and often has a high failure rate. As such, any technique that can minimize costs and time while also increasing success is considered the holy grail in biotechnology.

Coupled with advances in liquid handling, bioinformatics, and antibody expression and characterization, high-throughput surface plasmon resonance (HT-SPR) and single-cell B-cell receptor sequencing (BCR-seq) are driving the unprecedented scale, scope, and speed of biologics development. We will discuss antibody discovery by immunizing transgenic mice engineered to produce antibodies with human variable region heavy and light chains, and by using high-throughput B cell selection employing single cell sorting, Beacon®, or Nanovial Technology. These workflows maintain native pairings and output a diverse final set of binders encompassing a variety of clonotypes and heavy and light chain genes.

HT-SPR enables the evaluation of the largest of antigen–antibody interactions quickly and cost-effectively. It can also run parallel kinetic, affinity, and epitope specificity studies from the very start of the drug discovery process. scRNA-seq is a versatile tool for understanding cell types, cell states, and developmental trajectories in a diverse range of tissues and organisms. Increasingly, it is being harnessed to understand adaptive immune responses and facilitate a new paradigm of drug discovery.

During the webinar, our expert speakers will:

  • Share a robust method for antigen-specific antibody identification, including current industry advances
  • Help you understand cell types, cell states, and developmental workflows using BCR-seq or scRNA-seq
  • Discuss how high-throughput biophysical screening, epitope binning, and mapping are allowing therapeutic candidate selection from crude samples

0:00:07.3 Jackie Oberst: Welcome online listeners and viewers, and thank you for joining this Science AAAS Webinar, Discovering Antibodies in Months, New Advances in HT-SPR, single-cell BCR-seq and RNA-seq. My name is Jackie Oberst and I'm the Assistant Editor for Custom Publishing here at Science. I'll be moderating today's webinar. Joining us today as guest speakers are Dr. Derek Croote from IgGenix, San Francisco, California, and Dr. Lucy Liu from Alloy Therapeutics Waltham, Massachusetts, and whom we will learn more about shortly. Before we get started, I'd like to orient our online viewers to what's currently being seen. At the top right of your screen, you'll find a photo of today's speakers and a View Bio link, which you can click on to read more details about their background and research. To the right, you'll also see the resources tab, where you can find additional information about technologies related to today's discussion and a PDF of the slides. After the speaker presentations, we will have a short Q&A session during which we will answer some of the questions submitted by our live online viewers. So start thinking about some questions now and submit them at any time by clicking the Ask A Question tab. Also on the right, typing the question into the message box and then clicking submit.

0:01:15.7 JO: You can also log into your Facebook, Twitter, or LinkedIn accounts during the webinars, to post updates, or send tweets about the event. Just click the relevant icon at the bottom left of the screen. For tweets, you can add the hashtag #ScienceWebinar. Finally, thank you to Carterra for sponsoring today's webinar. And now let's begin. Although it is highly rewarding to create medicines, particularly therapeutic antibodies, doing so comes with several risks. It is expensive, time consuming, and often has a high failure rate. As such, any technique that can minimize costs and time while also increasing success is considered the holy grail in biotechnology. Coupled with advances in liquid handling, bioinformatics and antibody expression and characterization, high-throughput surface plasmon resonance or HT-SPR and single-cell B cell receptor sequencing, BCR-seq, are driving the unprecedented scale, scope, and speed of biologics development.

0:02:11.4 JO: We will discuss antibody discovery by immunizing transgenic mice engineered to produce antibodies with human variable region heavy and light chains, and by using high throughput B cell selection, employing single-cell sorting, Beacon or Nanovial technology. These workflows maintain native pairings and output a diverse final set of binders encompassing a wide variety of chronotypes and heavy and light chain genes. HT-SPR enables the evaluation of the largest of antigen antibody interactions quickly and cost effectively. It can also run parallel kinetic, affinity and epitope specificity studies from the very start of the drug discovery process. ScRNA-seq is a versatile tool for understanding cell types, cell states and developmental trajectories and a diverse range of tissues in organisms. Increasingly, it is being harnessed to understand adaptive immune responses and facilitate a new paradigm of drug discovery.

0:03:06.0 JO: Without further ado, I'm very pleased to announce the first speaker to you now, Dr. Derek Croote from IgGenix, San Francisco, California. As a co-founder and CTO of IgGenix, Dr. Croote is developing single-cell RNA sequencing technologies to advance our understanding of the immunoglobulin E biology and provide insights into the molecular interactions causing allergic diseases. His previous research in the laboratory at Stephen Quake at Stanford University focused on applying mass spectrometry, single-cell transcriptomics and next generation sequencing and bioinformatics to better understand a diverse set of diseases and and immunological perturbations including food allergies, flu vaccination, dengue infection, and glioblastoma. Thank you for joining us today, Dr. Croote.

0:03:56.4 Dr. Derek Croote: Thank you for the kind introduction, Jackie. As mentioned, I'm Derek Croote, Co-Founder and CTO of IgGenix, an early stage biotechnology company, developing biologics to treat allergic diseases like food allergy. Today I'll be talking about three things. First, I'll provide an overview of single-cell RNA sequencing. I'll then touch briefly on the process of antibody cloning and expression. Lastly, I'll describe high-throughput affinity and epitope binning using SPR and how we gain a comprehensive understanding of antibodies to the integration of these multitude of data. My talk today will be followed by Lucy Liu, our therapeutics and she'll provide additional insight into these and other topics.

0:04:40.7 DC: Before I begin, I'll say everything I mentioned today is work we do at IgGenix. Our process begins with blood samples of individuals with severe food allergies in which we use fluorescent activated cell sorting to isolate extremely rare IgE producing B cells. We then perform single-cell RNA sequencing to recover the full length paired heavy and light chain sequences that comprise monoclonal IgE antibodies. Lastly, we re-engineer these IgE antibodies that cause allergic reactions and allergic individuals into IgG4 based therapeutics that can block the allergic cascade. We'll start today with a fundamental question. Why study single B cells in the context of antibody discovery? The answer lies in the structure of antibodies as heterodimers comprised of two identical heavy chains and two identical light chains.

0:05:32.7 DC: Importantly, these two chains are encoded by distinct genes on separate chromosomes, and their pairing is what dictates specificity and function. Consequently, both methods that light cells and lose this pairing are incompatible with recovery of antibody sequences as illustrated by the cartoon on the right. Now, we've talked about why study single cells, but why use single-cell RNA sequencing specifically? I argue that with single cell RNA sequencing, you gain rich insight into not only the antibody but also the B cell that's producing it. So from single-cell RNA sequencing, you can get sequence and therefore specificity. You get gene expression, which gives you information on cell type, cell state and developmental trajectory. You also get information on gene splicing and variant calling, with some protocols and with others, you can also get information on cell surface expression, for example, protein markers, on the surface and fluorescent activated cell sorting.

0:06:36.7 DC: On this slide, I'm presenting just an overview of what single-cell RNA sequencing looks like. First, cells are isolated and we'll talk about ways in which they can be isolated. The next step is cells are life and the mRNA and those cells are reverse transcribed into cDNA. That cDNA is then amplified and then in a process known as library preparation barcoded in order to be compatible for next generation sequencing. On next generation sequencing instruments, typically Illumina machines this could be NextSeq or NovaSeq. These cDNA libraries are sequenced. The CNA sequencing rates are de multiplexed, and that data is analyzed, and we'll talk about a number of ways in which that analysis can occur.

0:07:27.3 DC: There are several methods of isolating single B cells. Early on, microfluidics was dominant, and this is a method that gives you really high sensitivity in terms of recovery of specific genes of interest but it suffered from lower throughput. Next in terms of higher throughput is fluorescent activated cell sorting, where individual B cells are sorted into individual wells of 96 or 3-4 wall plate. This method suffers a little bit from sensitivity compared to some microfluidic approaches, but you gain most importantly, the ability to sort cells based on various parameters. And in this way, you're able to positively and negatively select populations of interest and those that are not of interest. In this way, you can use fluorescent activated cell sorting to isolate extremely rare cell populations. And this is exactly what we do at IgGenix.

0:08:22.4 DC: We're after extremely rare IgE producing B cells and therefore need to enrich for this rare population and deplete other B cells that are less of interest for us like IgG or IgA. Lastly, in terms of the highest throughput approaches, you have droplets or micro-wells in which cells are encapsulated or physically slated into wells. This has really great throughput. You can get sequences for thousands, if not tens or hundreds of thousands of B cells. You do have some trade offs though, in terms of sensitivity or in terms of your ability to understand information about gene splicing or variant calling. But an advantage here is that there's lower expertise required, and a number of CROs even offer this as a service for those not ready to make an investment in the expertise or the capital required to bring single-cell RNA sequencing in-house.

0:09:27.5 DC: In terms of the molecular biology involved I can just caution in that there are now dozens, if not hundreds of single-cell RNA sequencing protocols, but not all are actually compatible with recovering monoclonal antibody sequences. And the reason for that is due to the way in which the mRNAs were transcribed and the cDNA is processed. So protocols that are compatible for antibodies involve those that produce, "full-length cDNA" through processes such as segmentation. And some of these include Smart-seq2 or 10X Genomics VDJ products. Protocols that won't work include those that are focused more on three prime and counting where the focus is on gene expression, and you don't recover the full-length sequence needed to bioinformatically reassemble into a complete contiguous sequence.

0:10:28.5 DC: How are those sequences actually reassembled? So as a starting point remember that NextGen sequencing on Illumina instruments produces paradigm reads that are quite short, so 2 by 150 or perhaps on the Myc code to 2 by 300, but that still is shorter than the contiguous sequence that comprise a heavier light chain. And so what's needed here is a process known as assembly. This is done bioinformatically and there are a number of software packages available that allow you to do this. But effectively what happens is there are overlaps in the nucleotides of various rig fragments that can be stitched together to produce this full-length contiguous sequence. And that's what I'm illustrating here on the right, in the various colors, you have kind of a purple V gene, a blue D gene, red J gene, and green constant region. And fragments from each of those distinct regions of the contiguous sequence are stitched together based on their overlaps.

0:11:34.9 DC: In terms of what we can do with antibody sequences, one of the really interesting things that I find, very powerful is looking at clonality of single B cells. This is a process in which you take the sequences and you group them by some common set of criteria. This is often the cells that or sequences that have the same V gene, same J gene, have the same CDR3 length, and have CDR3 sequences within some percent similarity. And this can vary based on stringency required in terms of determining ancestry, but effectively you're grouping cells that are related by the similarity of the antibody they're producing. And this is really powerful in terms of a number of interesting examples. One is detecting immune perturbations. So for example recent infection. Another is looking at antigen-specific antibodies that are blasting based on some sort of immunization or immune response. I studied flu vaccination and can say there was a large plasmablast population that emerged between day seven and day 21 after flu vaccination. And the cloning and expression of these antibodies from dominant clones at these time points produced antibodies that were specific to influenza antigens.

0:13:05.8 DC: Another useful approach or another useful example of clonality is characterizing B cell affinity maturation in which you can build phylogenetic trees that basically map over time in individuals if you're able to get repeated samples, the trajectory of the immune response and how the antibodies have affinity matured against the given antigen. Lastly, what you can do with clonality is understand convergent evolution. And this is something we do a lot at IgGenix. And here's just one example, is a really striking example where seven totally unrelated individuals have recombined or produced nearly identical antibodies that target an immunodominant allergen in peanut known as Ara h 2. In other words, individuals with peanut allergy are basically producing the same antibodies against the same antigens that causes disease. And so on the lower left, you can see that a lot of the somatic hypermutation of these antibodies is as we expect clustered in the complementarity determining regions or CDRs based on the high level of diversity.

0:14:17.5 DC: At those residues on the right, you could see a really interesting phenomenon whereby if you cluster the sequences by their similarity, their Levenshtein edit distance, you can pull out from this data the individualistic affinity maturation process that's occurred despite similar recombination in these individuals. So you could see on the side and at the top the cluster map shows colours that are grouped together. These colours represent the individuals of origin for these antibodies. And the fact that the colours are grouped together indicates that despite individuals recombining similar antibodies, it's an individual affinity maturation process. Not only can we understand a lot of information about the sequences of antibodies for example somatic hypermutation...

0:15:22.1 DC: Excuse me. Somatic hypermutation, clonality, a replacement mutation, silent mutation. But we can also overlay information about the gene expression of these cells in order to build a deeper understanding about the origins of the B cells from which these antibodies are derived. So I'm presenting just two examples here of two different disease states. The first on the left is the UMAP dimensionality reduction production from the Tabula Sapiens Consortium. This is a really kind of monumental effort by a number of groups in the single-cell RNA sequencing field and led by the CZ BioHub in California in which individuals donated organs effectively and groups associated with the BioHub Stanford UCSF, and Berkeley dissected those tissues, performed single-cell RNA sequencing in order to understand cell type and sub cell state in all tissues of the human body.

0:16:30.3 DC: And so I've just subsetted that in this example here. All the data is publicly available online by the way, just 264,000 immune cells and pointed out how you can look at the clustering these cells to identify cell state and B cells subtype. So on the left you can see clusters of Naive and memory B cells. And on the right I'm pointing out a cluster of plasma cells that originate from the bone marrow of several of these healthy donors. On the right is a different example rather than from a healthy individual, this is single-cell RNA sequencing and a UMAP projection for diffuse B cell lymphoma. This is an example where you can overlay information about the malignancy or B cell subtype as compared to healthy infiltrating cells and better understand disease state across individuals and within an individual over time. There's really power in integrating multiple orthogonal data here, not just on the antibody, but on the B cell of origin. I'll touch briefly now on the process of going from sequence to physical antibody and how to characterist those antibodies.

0:17:54.4 DC: So we talked about single-cell RNA sequencing as a method to discover the antibody. At this stage, after sequencing, you have the paired heavy and light chain variable regions. Those variable regions can then be synthesized. They can be cloned into heavy and light chain expression vectors, and those expression vectors can be co-transfected into mammalian cells for antibody production. In the biotechnology field, these mammalian cells are typically HEK293 or most commonly CHO cells. And many of the cell lines have been optimized for high titer productivity of antibodies. After these cells are allowed to produce sufficient amounts of antibodies secreted into the supernatant, you can collect that supernatant use protein A to purify the antibodies, and then elute the antibodies such that you now have the physical protein corresponding to the sequence that you discovered through single-cell RNA sequencing. In some protocols, you might already have specificity based on how you performed your single-cell RNA sequencing and isolation process. For example, with flow cytometry, you can do antigen specific-cell sorting or some micro-well, or encapsulation protocols allow the embedding of some specificity information or some readout of specificity at the single-cell RNA sequencing stage.

0:19:24.5 DC: But if the protocol you've performed doesn't have that component, the next step would be specificity screening. For us, our discovery platform known as the SEQ SIFTER Platform, is an unbiased one in which we're discovering all IgE producing B cells from the blood of individuals. And so at the time of sequencing and antibody expression, we actually don't know the specificity of the antibody. So we screen our antibodies against collections of antigens, in this case, food allergens, and other allergens in order to identify high-affinity antibodies that bind to allergens causing allergic disease. So once we're... Since we're able to confirm the specificity of our antibody, the next step in the characterization process is high-throughput affinity measurement and epitope binning by SPR, which would be the next step of this talk. I've shown in this slide some, at the bottom, some example timelines, just to give some general guidelines of what people might expect this might take. But these can vary quite substantially depending on the specific protocol, the scarcity, the sample and whether all of the infrastructure is in place to streamline these steps.

0:20:46.3 DC: Now, to talk about high-throughput SPR-based affinity measurement and epitope binning these are all data that we've collected and reports I'll present on the Carterra LSA platform. We have one in-house, and it's been really a workhorse for us in terms of identifying in determining the affinity of hundreds of antibodies against all different allergens that cause allergic disease in humans. And I'm presenting just a summary overview of that data here, rather than hundreds of binding curves. For several allergens, what we're finding from allergic individuals is that, the affinity maturation process that's produced IgE is one in which produced high-affinity antibodies to these allergens that are driving disease. So in the case of peanut, there's a couple allergens known as Ara h 2 and Ara h 6. In the case of shrimp allergen a dominant allergen is known as Pen a 1.

0:21:47.8 DC: This is Tropomyosin, a muscle protein. And then in cat, there's an immunodominant allergen known as Fel d 1. And this is a salivary protein that appears in cat dander. And in all of these cases, what we're finding from allergic individuals is that the antibodies being produced are extremely high affinity, some down into the single-digit picomolar range. So a really, I think, compelling example of using human samples as an origin for antibody discovery, when that's amenable to your target of interest. I'll next walk through the process of high-throughput epitope binning. This is a process in which of your first antibody, the one shown in black here on the left is functionalized or captured on the array surface. This is known as your ligand antibody. You then inject in antigen. It's shown in this as a red circle here.

0:22:47.4 DC: The last step is injecting your second antibody, a soluble antigen known as the analyte. And when you do this, a few things can happen based on the epitope specificity of these antibodies. In the first example, if the ligand antibody shares the same epitope as the analyte antibody, what will happen is the ligand antibody will block the binding of the analyte antibody, and you'll get no signal increase as measured by the centigram on the Carterra instrument. In contrast, if the antibody is bind to distinct epitopes, your ligand antibody will not block the antibody, the analyte antibody, and you'll get this effect known as sandwiching, which you can see as a very stark step signal increase in your SPR centigram. The power really comes from doing this pairwise competition across all pairs of antibodies in order to understand the number of unique epitopes on an antigen. And so here I'm presenting an example that likely many of us are familiar with at this point, this is for SARS COV-2 in which a consortium of individuals discovered antibodies through all different techniques or developed antibodies through all different techniques.

0:24:12.3 DC: These antibodies were measured for their affinity and also for their epitope bin determined by the Carterra LSA. And what you can see here is a result of that process, which is a cluster map of these individual antibodies in paralyze form. And so antibodies that block are shown in dark blue antibodies, that sandwich are in light blue. And from the clustering of these paralyze interactions, you can understand the number of unique, epitopes on an antigen and use that to basically in a process funneling and reducing the number of antibodies, you're struggling to select antibodies to proceed to epitope mapping. So really this is a nice step in reducing the collection of antibodies that you need to study down to a subset that bind to you to a unique epitope. For us, this is really useful in determining the immuno-dominance of a given epitope on an allergen. Not all epitopes are created equal and for us in terms of, that's especially true for driving allergic reactivity. So we use this quite extensively in order to determine to which epitopes majority of individuals with allergic disease are producing antibodies against.

0:25:32.2 DC: Overall, we've talked about kind of a three-step paradigm now, discover, in terms of single-cell RNA sequencing; engineering, basically identifying that sequence, producing that antibody. For us, we actually do a class switching step where we take IgE variable region sequences and graft those onto an IgG4 backbone. But in any case, it's the same process where you go from variable region sequence to physical protein that you then characterize. And we talked about, using the Carterra LSA to measure the affinity and look at epitope binning of these antibodies. But it's really the first step in several additional data that'll allow you to evaluate the functional properties of your antibody. I think now is a really exciting time in the field, given the rapid advancement of all of the aforementioned technologies. We're now at a point where for each antibody we can gain comprehensive insight into, not only their sequence, but the origin of the B cell from which it came. The B cell subtype, B cell state, we can tie in information about the specific organism, in our case it's donor clinical information or diagnostic information.

0:26:47.8 DC: And then use that data paired with functional evaluation of the antibodies to really select optimal drug candidates. And Lucy, will get into to some of these aspects as well, but ones I didn't touch on today include developability and activity. Ultimately, we wanna evaluate the functional ability of our antibodies in a given disease state to be kind of potent and active, and for developability. These are another orthogonal set of constraints that are applied to an antibody in terms of its ability to be manufactured. So, does it have a high melting temperature, aggregation temperature? Isn't sufficiently soluble, can it be produced? Meaning can antibody... Can CHO cells express at high titer? Is it pure in terms of monomer glycosylation pattern? That sort of thing, and is it stable? And so these are all important facets of ultimately selection of antibodies for or as drug candidates.

0:27:51.3 DC: So I'll end here, just by saying, we're performing everything I mentioned at IgGenix and attempting to build an unprecedented understanding of IgE biology on our path to developing allergy therapeutics. You can visit us on our website,, we're on LinkedIn, as well as Twitter, and there's my Twitter handle. And I'll lastly add that we're in the process of raising a Series B. So if anyone is interested in developing therapeutics or biologics specifically that really fill this unmet need in food allergy, please do reach out.

0:28:30.3 JO: Thank you, Dr. Croote. Our next speaker is Dr. Lucy Liu from Alloy Therapeutics, Waltham, Massachusetts. Dr. Liu is Senior Director and Head of Global Bio Analytics at Alloy Therapeutics, where she oversees development of comprehensive, lead selection enabling data packages for partner discovery services campaigns. She previously served as Director of High-Throughput Characterization at Compass Therapeutics, and prior to that as Project Lead at Biogen. Dr. Liu received her PhD in biochemistry from McGill University, after which she served as a postdoctoral fellow at Dana-Farber Cancer Institute, and as a postdoctoral scientist at EMD Serono In Vitro Antibody Technologies Group. Welcome Dr. Liu.

0:29:13.2 Dr. Lucy Liu: Thank you for the candid introduction. My name is Lucy Liu. So, I'm from Alloy Therapeutics. Thanks for organizer for inviting me to this webinar. So today I'm gonna talk about our workflow, in terms of the antibody discovery from transgenic mouse and PCR sequencing. So in this talk we'll cover humanized mouse model for antibody discovery, and then I'll talk about the high-throughput characterization for lead clones, and some of this data will be included in the next session, which is a binder recovery by B cell selection and sequencing. So it includes a B cell sorting, Beacon and Nanovial technologies. So at Alloy, we have a suite of transgenic mice... Mouse model that we use for antibody discovery. So, on the left here, it shows the transgenic mouse models that we call it the ATX-GK mice. And that with a different background, with a human kappa chain. So in all of this humanized mice, the antibody heavy chain and light chain variable domain are fully humanized. So, in this kappa chain, ATX-GK mice, it has a BL/6 mouse background as well as BALB/c.

0:30:42.5 DL: And we also have strains with the cross of those two strains. And we have to reduce the immunodominance, we have this real locus of heavy chain split into half, so the rest of the heavy chain variable domain and the light chain kappa chain is not affected. So we have these two strains, the proximal heavy chain and distal heavy chain. So each have half of this real locus of the heavy chain so that to reduce this immunodominance effect for some target. And to explore this, to have this diverse structure diversity and also sequence diversity, we also generated a lambda chain transgenic mice with these BL/6 mice background, mouse background. And also we have this hyperimmune transgenic mice. These are four strains with no culture or expression of certain genes to overcome the... There's some, limitation to generate of antibody discovery for difficult dark target. So, today all of these talks that we talk... In this talk, we talk about the result are generated from this ATX-GK mice on the left. So first I'll show you the validation data we have for this transgenic mice.

0:32:10.0 DL: So first we compare those mice to weight of mice in terms of the spinal size. So on the left you can see this total spinal size are comparable in the different strains of mice tested here. And we also tested for spleen B cells and the T cells. They're all comparable to all of these strains that we tested here. And next we test for plasma cells and both spleen plasma cells and the bone marrow plasma cells. And they show very comparable among those different strains. And lastly, we test for the secretion of IgG and the IgM. Testing are concentration in the serum, and they are also very comparable both under naive condition and also immunized condition. So we conclude that ATX-GK mice have a comparable immunophenotype to wild-type mice. So the next question is whether they represent the human antibody repertoire. So here on the top panel, it shows the frequency of V gene usage, in terms of comparing the ATX-GK cross mice that shows in blue to the human data. This human data is published by Vander Heiden in 2017. So overall we can see very good resemble of a human reference data for the V gene. And also here we have the J gene, the frequency also highly comparable of the two groups.

0:33:44.0 DL: And we also did comparison in terms of the identity. So that shows percentage of to the germline. So basically, showing this somatic hypermutation. And basically the profile of this ATX-GK mice is highly comparable to human data. So then after verification, we use those transgenic mice for our antibody discovery. So before talking about antibody discovery here, I include some of these categorization method we used for the antibody discovered. So as mentioned by Derek Croote, my co-presenter early, so he did a very nice summary of how this high-throughput SPR works. So here I'll just briefly mention that we can use it to evaluate purified antibodies in terms of kinetics affinity, epitope binning, and as well as these epitope for mapping by using peptide library.

0:34:49.8 DL: So I want to draw your attention in addition to evaluate purified antibodies, this method can also use for non-purified antibody in still in conditioned medium. So by screening soup, we can determine the kinetics affinity by using this capture strategy. So we can categorize for affinity, cross-reactivity, both species cross-reactivity and family member cross-reactivity. And also can characterize for blocking, for the ligand binding or blocking for these comparative antibody binding. So also we have cell binding assay to screen the antibodies in soups. And based on those data, we select a subset of antibody to purify for the final site of his. And for the purified antibody, we have a panel of assays that developed to evaluate those antibodies. For quality control, we use SECHPLC and CESDS, to determine the purity. And for DLS, for hydrodynamic radius and LC-MS for max ID, and also for PDM modification. For binding specificity and blocking as mentioned in the previous slide, that we can use high-throughput SPR.

0:36:09.4 DL: In addition to that we can use cell binding assay on the right, so we can determine by this titration determine EC50 specificity. By EC50 we can also do blocking, they gonna blocking IC on top of the cells. And those cells can be engineered cell lines or expressing the target or the cultured cell lines such as natively express the target. And we can also primary cells as well. We also evaluate the developability for potential developability of antibody as a therapeutic drug. So basically we evaluate the thermostability nanoDSF for turbidity. We test for self-interaction, polyspecificity or called polyreactivity to different reagents, and also we assess the hydrophobicity by HIC chromatography. So some of this data I'll include in later of this talk. So how we recover the binders after immunization. So this slide shows the platform we use for binder recovery. So we either use B cell selection or hybridoma fusion or phage display library basically is a immune phage library. We can also combine those method. So today I'm gonna focus on these B cell selection on the top.

0:37:38.5 DL: So, traditionally people use the antigen baiting for memory B cells, also we call the B cell sorting. Here we can incorporate with the NGS sequencing to have more diverse antibody discovered. And this, you can select a big number of positive hits, but if you want to include some of antibody characterization, you may consider on the bottom of the two technology. One is Beacon and the other one is Nanovial. So I'll talk about the workflows for these different technologies. So on this slide, it shows our workflow for single B cell sorting. So after on the top left, after immunization of ATX-GK mice, the immune tissue is harvest, we sorted antigen-specific memory B cells, and then is a single-cell RT-PCR. And then we can, automatically select his based on the sequence data. And after that is cloning and expression. In the middle of the bottom, it shows a high-throughput characterization, that's what I meant in the, few slide early that you can screen using the high-throughput SPR for the antibodies still in the soup.

0:38:58.2 DL: And we also do a cell binding for the soup screening as well. And then narrow down to a subset of antibody for purification. So on the left bottom, it also shows our NGS workflow. So basically we explore the repertoire and then, based on the select the B cell sequence that we can select additional likely binders from the NGS pool. So this slide shows one example. This is, target is TAA, Tumor-Associated Antigen. So we immunize with human antigen, human target. And here it shows the gating strategy. On left, it shows the selection of IgG positive live cells. In the middle plot, it shows these selection of B cells, basically IgG positive B cells. On the right, it shows these cyno target binders. So we immunize with human target and select here with cyno target. So basically we're looking for clones that cross-reactive to human and cyno of this target, of the target. And so on the right, it shows the distribution of sorted cells, from four different mice. And each mouse we have two immune tissues that are, spleen and the lymph nodes.

0:40:22.7 DL: So in total we have 1,500 cells selected by B cell sorting. And then, so on the bottom it shows the clonotype distribution, top panel shows the B cell, from B cell sorting, and the lower panel shows NGS repertoire. So here you can see that the clonotype distribution and mirrors in these, from the NGS data, this repertoire data. And based on the binders from B cell sorting, we can select additional clones from the NGS pool. So here the selected clones are purified, and here it shows iso-affinity plot to human antigen, on the left and to cyno antigen on the right. So here these control antibodies are color, shows in color, and that's over compared to antibody controls. So we can see that we generate antibody. We discovered antibody that can have tighter binding or similar affinity as or benchmark.

0:41:28.3 DL: And there's a lot of cyno-cross reactive clones discovered. And we also did the epitope binning to show this broad epitope coverage and this cell binding as it just showed this snapshot of one concentration that shows positive binding to the target expression cells, but not to the negative cell line. And in this slide, it shows convergence landscape to show this similarity plot. So basically from this B cell sorting, this project we discovered 88 cyno-cross reactive clones that representing 46 unique clonotypes. So next, talk about plasma cell screening using Berkeley Lights Beacon. So after immunization, plasma cells isolated based on the marker CD138, and then they're screened on Beacon for binding to multiple antigens for blocking and to like ligand blocking or compare the antibody blocking and for cell binding.

0:42:39.1 DL: And also importantly, we can include the functional screen on Beacon. So such as the reporter cell line as it can be used. And then the selected cells will be exported and sequenced and clone and for expression. So here is one example that, basically this is showing the assay principle that you use this multiple well, the chip, so basically you see the cells in those wells and the maximum you have one cells per well. And this well also called the PIN. So in this case, on the right side, you have this plasma cells, sitting in one well, and that it secreted the antibodies, most antibody and the B's that loaded with anti mouse antibody. On the top of this well that antibody combined to those Bs, the antibody are secreted from that plasma cell. And the green dots represent the fluorescence labeled antigen. So if the secreted antibody bind to the antigen, you will see the signal increase with time, so that we call blooming. So here it shows one example that, on the left there is human antigen assay, so you can see some of these active clones that shows in the yellow on the top of those wells, of those pins.

0:44:09.8 DL: And on the right side is IgG assay. So you want to select clones that shows [0:44:14.9] ____ in both assays. And then once you export those clones and sequence them, so here we have a plot of this usage of the clonotype. So on the x-axis it shows this usage of light-chain region and y-axis shows the heavy chain region usage. So this is a heatmap. So on the left, this heatmap represent these clones selected by Beacon technology. And, so blue it means that there's zero frequency and the warmer the color is the more the frequency is. So here it shows this clonotype distribution for the clones are selected by, from these plasma cells. To compare, we also plot the memory B cells for the same target from B cell sorting. So you can see the clonotype plot shows very differently on the two left plot and to compare them on the right... So on the right side, it shows combined of these two plot, basically if these clonotype used by both plasma cells and the B cells are showed in purple, if used by plasma cells only shows in blue, and if it shows used by memory B cells only, it shows in red. So you can see the pairing is very different between plasma cells and memory B cells.

0:45:47.8 DL: We also tested for CDR3 diversity. So here every clone in the set is compared to all the other clones in the set. And this is a heatmap. So blue, the dark blue means exactly same, the sequence CDR3. And then the warmer the color is, the more diverse they are. So the light blue means you need to change three or four amino acid to make the CDR3 sequence the same. And the yellow means you need to change 10 amino acid to make them the same. So you can... You know that the CDR3, the most abundance is like a 14-amino acid long. So you can see from this heatmap most the color is either yellow or red, meaning this is highly diverse for this CDR3 sequences. We also evaluate the binding affinity. So here it shows in blue the Beacon selected plasma cell the wrapped clones has a higher affinity than from memory B cells on the right. So in summary, that these shows, the plasma B cells that shows the heavy chain and kappa chain family pairing are very different from these memory B cells. And also plasma cell-derived clones shows higher affinity.

0:47:13.6 DL: So the next example case study I wanna show you is a defect called Target. So this is a TCR mimic project. So basically we want to select antibodies that bind to this HLA complex which are shown... The cartoon show on the top left in this box. So basically the red dots indicate that these peptide in graft on the top of the HLA complex. So the HLA complex, if they're using the same HLA allele, the HLA part of the heavy chain and the light chain B2M, they are exactly same if they use the same HLA allele. So the only difference is this peptide engraft on the top of HLA complex. And this peptide is about 8-10 residues. And then some peptide only differ by two amino acids. So it's kind of hard to select a specific antibodies. So that's why we have a few assays designed. So first on the left, under the bright-field image shows there's one cells per PIN per well, and then lower, and this on the left it shows IgG assay basically shows blooming of in the middle PIN shows that it's IgG secretion of plasma cell.

0:48:40.1 DL: And here basically it shows the assay where the proteins... So here you have this HLA complex, the positive complex shows on the top and the negative complex shows on the bottom. So it shows only reactive to the complex of interest. So next there we show on the cell, so basically this is a T2 cells, it add pulse on the red peptide or the control peptide CMV peptide. So here you can only see the blooming happens with this positive peptide. So this is the... From those assay it shows that the plasma cell on the second PIN, it is the right plasma cell we're looking for. So based on this data, we... Oops. So here the pie chart shows the percentage of positive clones that basically 22.8% shows specifically bind with the red correct positive peptide, but not to the negative peptide.

0:49:56.2 DL: And on the cells, we also about like a 22% shows to this, presenting this red complex HLA complex, but not the negative control. Interestingly while you compare this two group of clones, they only overlap about 30%, meaning if you only do this porting-based assay, you will have less than half shows the cell binding. And as the cell binding is more close to physical eligible condition, so it shows that is very important to include this cell-based assay for the selection. So now I show you two representative clones that discovered from this project. There are actually two different projects, but actually on the left it shows the specific binding from the cell binding assay and only bind to the cells plus with the Target peptide, the positive peptide, but not to any of those five negative peptides. And also it shows tied binding to those cells, present those the HLA complex. So it's either sub non-molar or single digit non-molar binding is 50. And also it shows concentration dependent binding to this HLA complex on the protein level.

0:51:19.9 DL: So as mentioned at the beginning, we have panel for assays to evaluated antibody hits. So here it shows that the purity is showing a good purity showing by CE and CSDS, good cyno stability is shown by this mounting temperature and T-ARC and good availability is shown by this polar reactivity. There's no polar reactivity. All right, I'll use the rest of the talk or just briefly talk about, these Nanovial-based antibody discovery workflow. So here the first step is cell loading. So nanovials those are small container, it's like a bow that, so here in the cartoon it shows the in the red, but in reality they are transparent nano gel. So in this case, we load the plasma cells on into these onto the nanovial. And here the nanovial inner it had internal, it has this CD138 to help reloading of plasma cells. And then we cannot remove the free cells after incubation for about one hour. This plasma cells start secreting these antibody and if the recognized antigen, so the antigen is already preloaded onto the inside of this nanovial, if it binds to this antigen, then it will tethered to the nanovial.

0:52:53.5 DL: If not then it will wash away. And after that is the staining part. Basically use this fluorescence label, the anti mouse IgG antibody, and then we can load to the sorter sought out into this microplate and for sequencing. The rest of these procedure of the steps would be the same as other procedures. So here it shows the gating strategy that you have this green gate showing the free cells. The red gate shows these empty nanovials. And the black gate shows the cell-loaded nanovials. And then we select by markers, by CD138 and also by ties the both markers for plasma cells. So now we gate on the plasma cell-loaded nanovials only. And if you have them at 4 degree which is on the top panel, you don't see any secretion of the mouse IgG. So they're only like about 0.4% background. But if you culture it 37 degree you are able to see some reactive to this anti mouse IgG antibody. And then you can sort those nanovials for sequencing.

0:54:10.4 DL: For this specific project, we sorted out 130 plasma cells and 62 we are able to recover the sequence and they are represented 53 clonotypes. So on the right it shows this clonotype distribution as well as these diversity of CDR3 sequence. So then we test for binding to protein there like a 37% binding to ECD monovalent ECD and 40% shows cell binding. And in the middle it shows cell binding curves. Based on that the EC50 can be determined. It shows on right, so here it shows the nanovial selected plasma cells. It has a higher tighter binding to the cells. Of course we cannot conclude from this one project, but it shows the red trend. And then we also did the epitope binning to show that the coverage of epitope in the middle of the two clones bind to a distinct epitope on the right to these two clones in the same PIN. And interestingly if you filter this, the cell binding result to this binning map, you can see the right PIN with the two red clones, they bind to the same epitope and apparently this epitope doesn't present on the cell surface, or it's buried in on the cell surface and not accessible because both clones shows no cell binding.

0:55:35.0 DL: All the other clones shows positive cell binding. And when EC50 can be determined, they are very tight binding. So sub nanomolar EC50 are determined. So in summary I've shown you we validate or transgenic mice, ATX-GK mice and we use that for our human antibody discovery and also I show you our B cell selection clone selection workflows including B cell sorting and also plasma cell screen with on Beacon and with Nanovial technology. So a lot of people contribute to this work. I want to thank everybody who contributed to generating data for this projects. So with that, I'd like to take any questions you may have.

0:56:33.5 JO: Okay, many thanks to Dr. Croote and Liu for their presentations. We are now going to move to the questions submitted by our online viewers. A quick reminder to those watching us live that you can still submit your questions by clicking the Ask A Question tab on the right, typing the question into the message box, and then clicking submit. The first question is for Dr. Croote. "Dr. Croote, how would you advise evaluating if SPR and HT‐SPR is right for a project?"

0:57:06.3 DC: I think over the last several years, the accessibility of single-cell RNA sequencing and high-throughput SPR has increased tremendously. And there are a number of CROs that are able to offer some of these services that allow you to pilot single-cell RNA sequencing on a given target. And I'd encourage that as probably a first step before making a large investment into the expertise and the capital equipment required to spin these programs up in-house. For high-throughput SPR though, and again I mentioned earlier, I think the Carterra LSA has been great for us, not just in affinity measurement, but epitope binning and guiding us towards optimal therapeutics in allergy.

0:57:49.5 JO: Good. Dr Liu, "How would you use your transgenic mouse models in bispecific or multi-specific antibody discovery?"

0:58:00.7 DL: So basically different light chain pairing is issue or difficult for antibody production for bispecific and multi-specific antibodies. So when we use antibody discovery, we can use... Currently we're using, common light chain immune phage libraries. So basically you prepare from this, tissues from immunized ATX-GK mice, so basically select the binders, antibodies against a different target from the same common light chain phage library. And another way we can do that is we also have a single-domain phage library, single-domain antibody. So basically if you use regular IgG for one target, you can use a single-domain for other targets, in this multi-specific, so you don't need to worry about light chain pairing. And in addition to that, for transgenic mice models, we are also developing the single-domain mouse model and as well as the common light chain mouse models. So our common light chain mice will be available early in 2023 and/or single-domain mice will be available later in 2023.

0:59:22.0 JO: Okay. Back to Dr. Croote. "When you reformat the IgE into IgG, do you silence the effective functions of IgG? As you know, IgG immune complexes can also lead to adverse effects."

0:59:35.2 DC: Great question. For the IgG subtype, we've chosen IgG4 for a couple reasons. One is because of the minimal effector function compared to, for example, IgG1. There's also precedent in this choice in the field of allergy, IgG4 increases by orders of magnitude over the course of immunotherapy. So nature is effectively telling us that in order to produce blocking antibodies over the course of an immunotherapy treatment, IgG4 can offer some of that protection. So we've chosen it for those two reasons. And then we also don't anticipate a lot of issues with complexes due to the fact that we expect to be in large stoichiometric access of our antibody on board in vivo, for individuals undergoing our treatment.

1:00:29.1 JO: Okay. Dr. Liu, "Can you comment between the Beacon and Nanovial platforms and give the pros and cons of each?"

1:00:38.4 DL: Right. So the good thing is that you can include some of these antibody characterization by both, using both method. And then in terms of Beacon is well established has been for quite a few years. So people using it, so then it's more very robust. But in terms of throughput, it is a relatively low comparing to a facts-based assays. And for Nanovial is a new technology that is... So it's a sorting based, so it is higher throughput. Of course some of the assays is under developing, like these sequencing, barcoding sequencing and also cell-based assays is now under developing right now.

1:01:32.3 JO: Okay. We have another question for you, Dr. Liu. "What are the throughputs of the B cell assays you talked about?"

1:01:40.1 DL: What is the... Sorry, could you repeat that question?

1:01:44.8 JO: Sure. What are the throughputs of the B cell assay you talked about.

1:01:48.9 DL: Yep. So the throughput like I mentioned early for B cell sorting and Nanovial technology, it used sorting method of fax method, it end for selection, so it's a higher throughput, so then we can screen millions of cells in a day well through for Beacon technology is relatively lower throughput, so it's about 20,000-40,000 cells in a day.

1:02:28.5 JO: Okay. Dr. Liu, again. "Are you able to rank the plasma cells for target binding affinity when you select them using Beacon versus Nanovial technology?"

1:02:38.2 DL: So unless you have IgG assay like for on Beacon assay, you can also use, for these stain IgG secretion. So you have some idea of the secretion level and then comparing to your target binding signal, you may have some idea in terms of a binding affinity, but usually there are a lot of variables. So we usually don't rank them for affinity. We basically just, for yes and no to select the cells. And then after that, when we express them, then we can do a more careful categorization in terms of ranking of affinity.

1:03:22.6 JO: Okay. Back to Dr. Croote. "What are the advantages of SPR over BLI?"

1:03:33.8 DC: I'd say for both you're able to get affinity of antibodies and both have been used extensively for that purpose. I see an advantage of SPR in the Carterra instrument in that it also has integrated ability to do epitope binning. And that's really powerful for understanding more context of the antibodies you've discovered, effectively, how many unique epitopes are there on an antigen, and what is the distribution of your antibodies to those epitopes? Those are all additional data you can get with this SPR design instrument.

1:04:08.0 JO: Okay. Dr. Croote, "How does developability of BCR-seq derived antibody look like? And how to improve it if needed?"

1:04:18.3 DC: For us we're doing all of our isolation from allergic humans, and so we have quite a bit of data now on developability of human BCRs from allergic individuals, and overall quite good. Our antibodies generally express very well. But there are some sequence liabilities. Just because the humans producing an antibody doesn't mean it's optimized for high titer stable expression in CHO cell lines at a CDMO, for example. So there will be some liabilities that often need to be removed in order to converge on a drug candidate that's optimized for the manufacturing process.

1:05:02.5 JO: Dr. Liu. "Do you ever combine facts with Beacon and the same workflow?"

1:05:09.7 DL: That's a good question. So, I think potentially that you can for this selection. But I am not for our current project we are not using combine of these methods.

1:05:30.9 JO: Okay. Dr. Croote. "Can you discuss the range of affinities found during your monoclonal antibody screening process?"

1:05:39.5 DC: I think those cells vary quite substantially based on the source of the B cells as well as the nature of the antigen or nature of the target itself. So it's hard to give a generalized answer. But what we've seen for human-drived antibodies the affinities can be extremely high, especially in the cases where someone has undergone some sort of immune challenge or immune perturbation repeatedly. And that's exactly the case with food allergies. The current treatment for many individuals with food allergies is avoidance in which you try to avoid what you're allergic to. And I can attest from personal experience having a food allergy that, that is ultimately unsuccessful. And so each accidental exposure is the opportunity for immune system to undergo additional affinity maturation, and produce high affinity antibodies. But high affinity antibodies is certainly not something that's limited to just human-drived antibodies. But there are lots of models and animals and strains that can produce high affinity antibodies that serve as foundations for drug candidate selection.

1:06:51.9 JO: Okay. Again, Dr. Croote. "Which script or program was used for BCR clonotype heatmap?"

1:07:00.2 DC: So I'm going to assume this is related to the conversion evolution slide where we looked at Levenshtein edit distance of heavy and light chain variable regions in individuals who produced antibodies specific to a peanut allergen. This was basically a script I wrote that uses SciPy as well as the Python package Seaborn for actually generating the cluster map and it has configurable options that allow you to add additional metadata to the antibodies. And in the case that I presented individual origin was one of those metadata that showed an individualistic process of affinity maturation.

1:07:46.6 JO: Okay, Dr. Liu. "How do you choose whether to run the Beacon or Nanovial process for a plasma B cell workflow? Or do you typically run both in parallel?"

1:07:57.1 DL: Okay. That's a good question. It really depends on these the target, for some target, like for... Depends on the essay that we wanted to do. Like for example, if you have a reporter cell line, so it's more established on Beacon that to include those functional assay. Of course, as I mentioned earlier, that Nanovial so pertaining about signs, they also developing cell-based assay as well. So then it depends on the target. And also I want to comment on. So it's related to a early question about like a combined facts assay and with the sorting essay with these Beacon assay. So, it depends. What do you meant combined? So I already, I understand it as like to evaluate the same side of the binders. But actually for different target, we actually use a different method to select the binders. So definitely for the selection, we have a lot of cases. We use both Beacon to select the binders, and we also use the fact-based method to select binders as well. And as shown in one of my examples that they can cover different pairing of this different pairing of the heavy chain and light chain and then shows a different, a more clonotype coverage.

1:09:27.2 JO: Okay. We've come upon the end of the hour. It just remains for me to thank today's speakers, Dr. Derek Croote from IgGenix, San Francisco, California, and Dr. Lucy Liu from Alloy Therapeutics Waltham, Massachusetts. Please go to the URL now, at the bottom of your slide viewer, to learn more about resources related to today's discussion and look out for more webinars from Science available at This webinar will be made available to view again as an on-demand presentation within approximately 48 hours from now. We're interested to know your thoughts of this webinar. Send us an email to the address now up in your slide viewer, Again, thank you to our speakers and to Carterra for their sponsorship of today's educational seminar. Good bye.