At Twist Bioscience, our antibody engineering platform has revamped the discovery workflow to take advantage of the ongoing revolution in synthetic biology. In six weeks, we designed, produced, and screened a new antibody library and identified high affinity IgG and VHH candidates to the S1 spike protein of SARS-CoV-2, as well as to the human ACE2 cellular receptor to which SARS-CoV-2 mediates viral entry. Crucially, the high throughput and low reagent use of the Carterra LSA platform has allowed us to take full advantage of our antibody expression and purification workflow to characterize hundreds of the lead candidates by affinity, specificity, and epitope bin.
0:00:00.9 John McKinley: Welcome everybody, and thank you for attending today’s webinar. During the webinar, please send your questions by private chat to me, John McKinley, the host, and they will be answered at the end of the webinar. Carterra is helping many of our customers screen libraries of antibodies in the race to find cocktails of therapeutic antibodies to combat COVID-19. We also have a new nickel biosensor chip for the LSA here at Carterra. And today, we’re fortunate to have Tom Yuan from Twist Bioscience present on success of their antibody engineering platform and workflow in their fight against COVID-19. Tom received his PhD from the University of California Irvine under the mentorship of Professor Gregory Weiss. Tom was an early scientist at Surrozen, where he developed bispecific antibodies as Wnt agonists in the regenerative medicine space. Tom then joined Twist Bioscience in 2018 as part of the Biopharma vertical to drive the antibody discovery and optimization by leveraging Twist’s unique DNA synthesis platform. Tom, I’ll hand it over to you now.
0:01:30.4 Tom Yuan: Thank you, John. I hope everyone can hear me nice and clear. I’d like to thank Carterra for inviting me to this webinar and of course, everyone for spending their time today to view this webinar. Today, I’m just gonna be going over some of the preliminary data that we have in some of the drug antibody drug discovery efforts that we’re looking at towards for fighting the COVID-19 pandemic. Obviously, this has had a huge effect on all of our lives and it’s something that we’d really pivoted towards in the last couple of months. Admittedly, we’ve actually gotten quite a late start to this. I’ll go over this in the webinar, but everything that you see here is actually compressed to about a six-week timespan from the creation of some new libraries specifically targeted towards anti-SARS-CoV-2 S1 spike protein all the way to reformatting to four IgG to some of the initial kinetic characterizations that we have here. Twist, if you are not familiar with Twist, Twist Bioscience is the parent company. Their main thing is with DNA synthesis. So their core technology is basically the silicon chip platform where they’re able to make up to 96,000 genes on a single chip. Typically, if you do oligonucleotide synthesis, it takes an entire 96-well plate to make one gene. At Twist, this scale is what you know several orders of magnitude higher, which is why they’re able to provide many of the oligonucleotide orders clonal gene orders at a much cheaper price.
0:03:18.9 TY: The department that I’m in is called Twist Biopharma and Twist Biopharma is essentially started under Aaron, the mentorship of Aaron Sato mainly to leverage this DNA-synthesis platform for antibody discovery and antibody optimization. So the primary way that we use this is we will use the… Print the oligonucleotide pools to create antibody libraries. So for example, in a typical antibody, let’s say in the heavy chain, you have the constant framework regions in green and then the invariable CDR regions in these other colors. Typically, the way we design these libraries is that we specify the exact CDR sequences that we want to insert into each domain and then these get mixed combinatorially so that we have… We normally try to target within variable heavy or variable light, about 10^10 in terms of the theoretical diversity. This contrasts to a more traditional mutagenesis approach when using NNK or TRIM to assemble your antibody libraries where you can’t precisely control the exact sequences that are going into your library, when we design our libraries that are either fully synthetic or semi synthetic, you print the exact sequence that you want for CDR1 and then you can designate exactly how many… How many of those CDR1s you want. They can also vary in length.
0:04:58.0 TY: And we know exactly what’s going into the library. And we also NGS sequence our libraries to make sure that the output of the library once we’ve constructed it exactly matches the initial design that goes into it. So you can rationally sample your desired sequence space. Obviously, when you’re printing the exact sequence, you can remove any amino acid liabilities such as isomerization sites, cleavage sites, deamidation sites, glycosylation sites so the initial library that you’re putting into does not have these to begin with. You can also do things like include specific motifs that you want or exclude specific motifs that you don’t want in your antibody library. So what does Twist offer with this silicon-based DNA synthesis? Like I mentioned before, you have exact… You can avoid specific restriction sites. You can vary the length of your CDRs. You can create libraries that specifically ratio controlled the amino acid distribution at each site. Alternatively, you can literally select the exact sequences that you want to put in as well. And you can also include multiple germ-line scaffolds and like I mentioned, we verify our libraries by NGS.
0:06:10.1 TY: So the way that we’ve been tackling both this project and also other projects, is that we don’t use a single library. We use multiple libraries to target any either soluble antigen or cell surface antigen. We have specific libraries that we have developed tailored around targeting GPCRs, targeting ion channels. We also have another separate platform to optimize and humanize what we call TAO, The Twist Antibody Optimization Platform. And we’ve actually… As we’ve developed the lab, we’ve also generated… Our Protein Sciences team has done a great job and generated a new workflow for high throughput antibody production. And what this allows us to do is once we have the initial leads and our phage display libraries, we can convert or reformat through IgG hundreds of antibodies at a time.
0:07:08.9 TY: And we typically do this at a 1.2ml scale, and that actually gives us enough protein to do the initial kinetic characterization, some epitope binning and even some computation data just from that 1.2 ml transient transfection. And this actually plays well with Carterra as well, because the Carterra is a great system for assessing hundreds of antibodies at a time, but the way it prints these antibodies is great because you can have a very low concentration, say five or 10 micrograms per ml. And you actually recover most of your antibody when you print them onto the chip. Which is actually great because once we print them on the chip, we can use the chip over and over again to assess kinetics, we can use it to assess epitope binning, but then still retain that IgG and do some other downstream assays.
0:08:11.5 TY: So this is just a quick slide kind of overviewing some of the libraries that we have on hand right now. One of our big focuses is with actually single domain VHH libraries. We have several different forms of this. We have libraries that are targeting ion channels, carbohydrates, GPCRs, and also libraries derived from what we call Hyperimmune libraries where the CRH3 will have actually millions of different sequences. So what’s a little bit different about Twist is that we can actually generate these libraries and have, make these fit-for-purpose libraries within an extremely short amount of time. And also the discovery cycle is fairly short as well.
0:08:52.4 TY: So our typical workflow is around eight weeks. After we design the actual, the antibody libraries, we are primarily putting them into a phage display vector. So in this case we’re displaying small fragments of the antibody on bacteriophage, and the bacteriophage gives us that… The vector phage is what we’re using to actually pan against or look for binders for a soluble protein or cell-based targets. This typically takes about two weeks, and once the panning and screening is done, we sequence all the positive hits that we have, either by NGS or Sanger depending on whether it’s a soluble protein or if it’s a cell-based target.
0:09:42.1 TY: It then takes roughly about three weeks to reformat to an IgG and DNA scale up. So this is a little different because we do, do some initial hit-picking in the phage display format, but all of our screens are primarily after reformatting to the four IgG because obviously we will not wanna test the final product, we’re not testing antibody fragments, we’re testing the final product after it’s been reformatted to, say, IgG1 or IgG2 or VHHSC. This then goes through high throughput IgG purification, it takes roughly a week to do this. And then from this we use this plate to conduct the binding functional assays.
0:10:31.7 TY: Typically for our high throughput IgG, for a 1.2ml scale we get roughly 100 or 200 micrograms of usable protein or usable antibody, which is enough to conduct the initial kinetic characterization to the Carterra LSA.
0:10:52.1 TY: This slide is going over some of what we call… Our high throughput antibody protein production. We’re actually planning on introducing this as a new alpha product where you’re able to actually submit your antibody sequences that you want to express and purify, and we will go ahead and express these for you at either the 1ml scale or 8ml scale. Purify them by protein A and deliver that purified antibody in either a glycine buffer or an amine-free sodium acetate buffer. So we can do this with IgGs, single domain antibodies. Obviously, Twist already delivers purified DNA, but we’re able to supplement this with the purified antibody, and like I mentioned, this is a Thermo Expi293 transient system.
0:11:44.1 TY: Some of the metrics we can apply, introduce as well are… Provide as well is digital SDS, percent purity and yield and some additional optional tests as well, including affinity assessment by Carterra. So looking at the COVID-19 pandemic, one of the recent publications, I’m sure if you’ve been following this, you’ll see that ACE2 is our human angiotensin-converting enzyme 2 is one of the primary cell surface receptors by which SARS-CoV-2 mediates cell entry and infection. TMPRSS2 is also required, this is a serine protease that cleaves, I believe the S spike protein which then mediates the infection by the VH2 surface receptor.
0:12:40.5 TY: But this paper actually looks at how SARS-CoV-2 is phylogenetically related with some of the other known Coronaviruses. So if you look at by BI phylogenetic tree, you see that SARS-CoV-2 clusters with the S proteins of known bat coronaviruses. And then if you can look at the actual sequence of SARS-CoV-2 to the spike protein, the spike protein S1 tier, is gonna be tier, RBD is the receptor binding domain, RBM is the receptor binding motif and they’re able to see just by a sequence alignment that ACE2… That it’s highly similar to the bat coronavirus S1 proteins that are able to bind ACE2, and there’s a low similarity with the bat coronaviruses that are not able to bind ACE2 these two, so this is sort of the first indication of what the actual cell surface receptor was that mediated viral entry.
0:13:47.8 TY: This is another paper published in Cell that actually looks… That actually looks to see if the introduction of soluble ACE2 actually reduces the amount of internalization that you see from the SARS-CoV-2 virus, and they’re able to show that, and when you introduce soluble human recombinant ACE2 in solution that you’re actually able to lower the SARS-CoV-2 internalization which suggests that human ACE2 is the primary receptor by which the coronavirus mediates cell entry. So we’re able to use this and try to tackle and reduce SARS-CoV-2 infection by two different mechanisms. One is by targeting antibodies against the S1 protein itself on SARS-CoV-2, the other is actually looking to see if we can block viral entry by blocking the ACE2 receptor on cells and by blocking the ACE2 receptor, it should be able to block the binding of SARS-CoV-2 to ACE2 and then thus decreasing the amount of virus and viral internalization.
0:15:09.1 TY: So what we’ve done is… So actually before then this is just a cryo-EM we’re looking at, the pre-fusion complex of the spike protein, and if we look at exactly where the spike protein binds to ACE2, we actually see that the RBD domain actually binds away from the catalytic site of ACE2, which is important because we wanna make sure that… So, we wanna make sure that we’re not affecting the catalytic site or possibly not affecting the catalytic site if we design antibodies that block the S1 RBD binding interaction with human ACE2. So, we’ve also announced the collaboration with Vanderbilt University, where we’re taking… We partner with them to design a synthetic library based on sequences from a convalescent COVID-19 survivor. The idea behind this is that if you have a COVID-19 survivor already generating antibodies, we’re able to take some of those sequences and then design a new semi-synthetic library, based off those sequences to potentially target and more effectively target the SARS-CoV-2 S1 spike protein.
0:16:34.7 TY: So some of the libraries that I’ll be showing in this specific webinar actually have sequences derived from this library. This slide is going over the general overview of the targets that we… In terms of the entire workflow, so the two targets that we’re looking at in terms of finding antibodies to bind to ours, the SARS-CoV-2 S1 RBD domain and then also the ACE2 ECD. We’re using multiple different libraries of both scFv format, VHH format, and each of these libraries are over 10^10 diversity each. In terms of a phage just like panning, we typically pan about three to four rounds, what this means is we’re taking these antibody libraries, exposing them to either SARS-CoV-2 S1 or ACE2, allowing them to combine and then washing away the non-binders, taking that pool and enriching that pool, and then repeating that three or four times over.
0:17:39.2 TY: And the idea is each time we repeat the rounds of the panning that you will go from a large billion member antibody library, and by around three or four, you’re narrowing that antibody library to thousands of clones, but you’re enriching the amount of binders that actually bind to your antigen of interest. Typically, we screen for about 1200 separate colonies for each panning round, and to make sure we capture as much diversity as possible, and then we go into the colony expression, and like I mentioned, we use our… The high throughput IgG product offering, the Alpha product offering actually internally for ourselves as well, to produce enough antibody for kinetic characterization. And for testing, we look at Infinity, we look at the receptor looking competition, and also we look at the viability to make sure that these antibodies are not… To make sure these antibodies are, for example, thermostable… They’re not sticky, they’re not… They Don’t bind to everything, they see specific binding to either ACE2 or the SARS-CoV-2 S1.
0:18:51.1 TY: And like I mentioned, we try to compress this time… That timescale where we’re going from a library design to actually having some initial hits in hand or in about a six-week timeframe. Obviously, if we had to skip… We also threw in some existing antibody… Other antibody libraries, not just the CoV-2 phage display library, but also our hyperimmune phage library, and also our single-domain phage libraries, to see what we can extract out of these.
0:19:25.3 TY: This slide is showing the phage selection in rounds and titers. Like I mentioned before, when you’re dealing with phage display campaign, the goal is to enrich for antibodies that bind to your antigen of interest. So in the first initial rounds, you want to be gentle with your stringency so that you don’t wash away potential binders that have, say, very low copy number in the first round. As you go through the rounds, you want to increase the amount of washers, so you increase the amount of stringency and you’re lowering the amount of antigen that’s present, so that the… You’re further enriching for productive binders.
0:20:13.4 TY: The numbers here in bold for the different project codes indicate the phage titers, so what this is showing is that as you’re going through each round, you want to see this phage titer slowly increase. Basically, that… It gives you a first initial peek as to whether your phage selection is actually working or not. If you’re not seeing them increase, typically what you’re doing is you’re enriching for non-functional binders, or you’re enriching for antibody clones that happen to express faster than other clones, but if you see this nice increase, say for example, from two times ten to the sixth to 1.4 times 10 to the eighth, then, this is a strong indication that you’re giving productive binders to the antigens that you’re… Of interest.
0:21:02.4 TY: So this is some of the phage ELISA data just demonstrating that with these antibodies expressed on phage that they actually bind to either SARS CoV-2 Spike S1 protein or the ACE2 protein. And you can tell that from… Just screening the ELISA data that we’re getting hundreds of potential hits in terms of looking at the OD fold-over background. So what we did was we took all these ELISA positive hits, sent them in for sequencing, and once they were sequenced, then we reformatted them into a full either IgG-1 or the S1… Anti-S1 antibodies or an IgG-2 for the anti-ACE2 antibodies.
0:21:56.8 TY: This is looking at just the phylogenetic tree, looking at both the ELISA data, and we did… Also did a crude binding inhibition study looking at these anti-SARS COVID-2 S1 binders and whether they’re able to inhibit binding to ACE-2. This is a rather crude assay, but it gives us some initial hint before we spend the three weeks converting these antibodies to see if they have any potential to block the S1 ACE2 interaction. What you’ll see here is that once… Bars that are deep blue show that they have potential high binding inhibition with ACE2.
0:22:42.3 TY: If we reverse this assay for the anti… ACE2 binders, we see that some of these clones as well have high inhibition binding, or they… In addition of the phage, gives you high inhibition of the ACE2 S1 interaction as well. And by looking at the phylogenetic tree you see that this… The binding inhibition does not cluster around a lot of this highly similar family of antibodies there. We actually see a nice distribution in terms of these… Potentially… These antibodies that potentially inhibit the S1 ACE2 interaction.
0:23:25.6 TY: So, we also didn’t… Before we converted this to the full IgG, we have to do a specificity study looking at where exactly on S1 they did bind. We see that we have a nice distribution of some antibodies that bind to S1, some bind to S1 specifically the RBD domain, and actually, in our case and almost none of these bound to the RBM domain itself.
0:23:52.2 TY: So, for… Where the Carterra LSA really comes in is because we already have such a high throughput workflow where we’re not picking, say the top 10 or top 20 leads, we’re taking 100, 200, 300, 400 antibodies at a time and converting all of those. The LSA itself actually is really useful for us because it allows us to immobilize and print up to 384 antibodies onto a single chip at a time.
0:24:21.5 TY: So, on the chip itself, the Carterra LSA has two main printheads. There’s a 96-channel mode and a single channel mode. So like I mentioned, with the 96-channel mode, this is how you’re actually… In our case we’re printing the antibodies on the chip in the 96 channel mode and printing them in a grid of four so that you get a total of 384. Once the antibody is immobilized on the chip, then we switch into the single channel mode and then flow over say… The S1 protein or the ACE2 protein usually with the increase concentrations, to get the actual kinetics… Receptor binding kinetics.
0:25:07.4 TY: In terms of the LSA platform’s core applications, we’re mainly using the kinetics and affinity. I don’t have the data yet in this bit of the webinar but we are planning on running a full epitope bin on all the anti-S1 and anti-ACE2 proteins. And the LSA can also be used for epitope mapping and quantitation as well.
0:25:28.2 TY: In terms of the 96-well printhead, typically what it will do is, you have a plate, the 96-well printhead will take up your antibody and then flow into the chip, back and forth. So what’s nice about this is that, say, if you have a low concentration of antibody, say 5 micrograms per ml or 10 micrograms per ml, because it’s flowing the reagent over back and forth, left and right, it’s not… You can increase the amount of time in which you’re allowing the antibody to get mobilized without increasing the volume that you need. So typically for this, you need about 200 microliters to do the print, but your concentrations can be fairly low and you can just increase the amount of time to immobilize, to increase the total amount of antibody that’s bound to the chip.
0:26:24.6 TY: In terms of the single-channel flow side, so once the antibodies are actually coupled to the chip, the single-channel flow mode is used for flowing over the same antigen over the entire chip at once, so if you have say your spike protein you’re looking at real-time binding for every single spot, all 384 spots on this chip at one time.
0:26:53.6 TY: So getting to some of the data. So it looks like we have actually a good number of picomolar level and low nanomolar affinity binders for both ACE2 and also S1. So for ACE2 many of these are 0.2, 0.3, 0.7 nanomolar. For S1, there’s a few that are picomolar level but there are many that are low nanomolar. And if you look at the binding kinetics, this is also very apparent where you see, you know, a nice Kon and then you see a very, very slow payoff Koff for many of these clones. We also tested this against… We tested this against S1 but we also tested against a full length S-Trimer protein that ACRO offers as well to make sure that it actually binds to the trimeric form of the full S protein.
0:27:54.2 TY: And we see that in many cases, that the antibodies that are able to bind the S1 monomer are also able to bind the full length S-Trimer protein. Obviously, with these, you see that the k off are even slower but this is no penetrability trimeric complex.
0:28:18.3 TY: So, conversely, we also see that for the ACE2 binders, we have many picomolar level binders as well, for example through the 1A3-3, we see a KD of around 326 picomolar. Some of these… We also have a nice distribution as well, so they’re not all picomolar level binders, there are ones that are a little bit weaker, single digit nanomolar, double digit nanomolar but this gives us a nice kinetic diversity in terms of trying to develop these antibodies.
0:28:49.7 TY: One of the things we’re actually also worried about is that ACE2 can be a rather sticky protein to assess, so we wanted to make sure that in our assay that it doesn’t bind to say a non-binding antibody, so we put in Trastuzumab just to confirm that the ACE2 binding that we’re seeing here is not from some stickiness where the ACE2 is just binding to the chip overall. And this is also obviously apparent because it’s not… The binding kinetics are not exactly the same for exo antibody.
0:29:31.7 TY: This is some additional data, looking at just some more clones, looking, you know, that have very low picomolar, or low single digit nanomolar binding to either S1 or the S-Trimer. And one more thing we actually also tried is actually trying to see if we can use the Carterra LSA for an inefficient screen. So typically in this format, you can immobilize the S1 protein onto the chip, and then if you have just the ACE2 protein in solution flowing over the chip, you’ll see that it should be able to bind S1 and you see that with these curves here, so these curves here, this thick band here are the ACE2 binding to immobilized S1.
0:30:22.0 TY: Many of these are antibody plus ACE2 premixes showing no effect on the ACE2 S1 interaction. And then there are a couple of clones here that show very clear inhibition, where when we mix in in molar access the antibody with ACE2 that it completely blocks binding to the S1 protein on the chip circuit. We’re not using this with a primary screen, but because we already have the chip… We already have the antibodies coupled with the chip, we use that chip to assess kinetics, it’s pretty trivial to also couple S1 to the same Carterra LSA chip and then flow over ACE2 as well.
0:31:03.0 TY: So we’re able to do the kinetics and also the initial inhibition screening in one go. So this is actually kind of a neat, sort of neat use case that we have for the Carterra LSA. And we actually went back and redid this inhibition assay in an ELISA format as well just to see if there’s a real… If our antibodies are truly neutralizing.
0:31:31.6 TY: So for the S1 RBD ACE2 inhibition study we used an existing kit from ACRO, in a computational ELISA format. What this is showing on the left is essentially binding of Biotinylated human ACE2 to immobilize SARS-CoV-2 S protein on a MaxiSorp plate. So essentially you’re mobilizing your S protein on the plate, and then you’re introducing human ACE2, that’s Biotinylated, and then you detect that binding via a Streptavidin-HRP. So you’re detecting specifically the Biotinylated ACE2 binding to SARS-CoV-2 S protein on the plate.
0:32:17.2 TY: When you introduce an antibody in manner, as you increase the amount of antibody, you should be able to see decrease in the amount of Biotinylated ACE2 that you see bound to the ELISA plate. So at low concentration, for example, there’s not enough antibody to block the interaction, so you see almost all the Biotinylated human ACE2 binding to SARS-CoV-2 S protein on the plate. At high concentration of the antibody, you see this stiff decrease and with the proper dose response, you see a nice S curve and you’re able to extract an IC50 out of this.
0:33:05.8 TY: So we did an initial ELISA competition experiment looking at some of our proteins here. It’s a little messy because we’re showing quite a lot of these antibodies, but what we’re looking at here is essentially on the left hand of the X scale is a low concentration of antibody, almost a 0.1 nanomolar, and then at the right… Scale of the X-axis is high concentration of antibody, over 100 nanomolar. And what you see for some of these is that you have actually a very good dose-dependent manner in terms of which you’re inhibiting that interaction of the Biotinylated ACE2 in solution binding to the immobilized S protein on the ELISA plates.
0:33:47.0 TY: So for example, there’s… ACRO also provides a neutralizing antibody that’s shown here in these circles here, and you see that some of these anti ACE2 antibodies are able to actually improve upon that and will give a lower IC50 compared to the ACRO antibody. This is also another set here where it’s showing some of these antibodies… Anti ACE2 antibodies have… They drastically inhibit that ACE2 S1 interaction here with very low percent down to almost to zero, and of extremely low IC50 for many of these anti ACE2 antibodies.
0:34:40.5 TY: We’re also planning on testing this with the anti-S1 antibodies as well, I’m probably mainly showing the anti-ACE2 antibodies here. One thing to note is that in the Carterra experiment that I showed with… The competition experiment, some of the main ones that were inhibiting were 1A4-7 and 1A4-57, which are also shown to be highly inhibitory in the ELISA format as well, so, it’s nice to see those two data correlate.
0:35:23.7 TY: So it terms of our lead summary, we have… Within six weeks, we have totally reformatted some of these IgG leads using our High Throughput IgG product offering for anti SARS-CoV-2 S1 protein we have 59 leads, actually for the first set, we are… We potentially might have more as well. For ACE2, currently we have 184 leads both in IgG2 format and VHH-Fc formats. And we have a nice… We have some high-affinity picomolar level binders, but… And inhibitors, but we also have a nice spread as well, so they’re not all on picomolar level.
0:36:01.1 TY: And in terms of future activities, we wanna test all these antibodies, fully test to make sure that they blind every… Block that receptor leaking interaction. And we also want to conduct a full epitope binning using the Carterra LSA platform. We wanna make sure we test some of these tetravalent and monospecific and bispecific constructs, and also obviously do a full developability assessment. We wanna be able to partner up with CROs to do in-vitro functional testing, so the pseudovirus reporter assay and also viral neutralization assays and long epithelial organoids, and like I mentioned, we do have additional VHH screens and SARS-CoV-2 S1 protein.
0:36:47.7 JM: So to get it into… So we’re still continuing potential screening for these antigen leaks. So, in terms of ways to work with Twist Biopharma, we’re not a drug development company, but we do work with many collaborators to either discover or optimize antibodies, so the slides shows you many different ways you can work with Twist, either with licensing libraries or partnering with us during new leads or partnering with us to optimize the existing leads. And like I mentioned are the High Throughput IgG production of the new alpha product that we’re starting as well, and we used this partly in house as well to be able to produce all the antibody protein that we need.
0:37:47.2 TY: And with that, I’d just like to thank everyone on the team, this has been a huge effort, a really fast pivot and a huge effort, both within the Twist Biopharma department, but also with the library’s team and protein science team, and many of the different functions at Twist that without them, we wouldn’t be able to either design and synthesize these libraries or test the throughput that we’re able to do. And with that I will take any questions you might have for this webinar as well Noah is able to answer any technical questions you may have about Carterra. And thank you for the time.
0:38:34.2 JM: Thank you, Tom. It was an excellent presentation, very informative. We will now open up a panel discussion with Tom and Noah Ditto, Carterra’s technical product manager. Noah supported drug discovery and early clinical development for nearly a decade at Bristol Myers Squibb. He is an expert in label-free biosensors screening technology, and is one of Carterra’s longest term employees. If you have any questions, please send them by private chat to me, John McGinley, the host, and we will answer them. Okay, I see some questions coming in, one of them is regarding the data, how you use the data from convalescent patients to create libraries, and how many of your binders came from those libraries compared to naive libraries.
0:39:27.3 TY: So I can’t get into specific details, but the essential idea is we take the antibody sequences from someone who has recovered from COVID-19, and we can take some of the CDRs into… And incorporate those CDRs into our synthetic libraries in a combinatorial fashion. In terms of the percentage… Actually, most of our anti S1 hits came from that library, so I would actually say the majority came from that library.
0:40:09.6 JM: Another question, Tom. As a user of other kinetic platforms, how’s the Carterra LSA differentiating for you at Twist?
0:40:19.7 TY: So the LSA has actually been pretty instrumental in terms of our workflow, typically with some of the other label-free SPR methods, we actually do have another smaller SPR that can only run two channels at a time, but we only use that for just verifying binding for something we already have, the LSA has actually allowed us to screen kinetics for IgGs and get real-time kinetics as a screening tool, not just the tool to confirm binding, say after we’ve done a lot on the ELISA screen or after we’ve done a Fag screen. So it’s completely revamped our workflow.
0:41:06.0 JM: Tom, what regeneration condition was used for the SARS-CoV-2 kinetics?
0:41:14.5 TY: The regeneration in this case, actually, we’ve been using the Thermo IgG Elution Buffer, which is commercially available. So you can use that, we’ve also used glycine in the past as well, glycine pH 2, it seems to work well and it doesn’t… At least in our case it does degrade binding of the antibody to ACE2 antibody S1 interaction.
0:41:36.1 JM: Thank you, Tom. We have a question for Noah. What are some of the considerations when epitope binning antibodies against the spike protein?
0:41:48.3 Noah Ditto: Oh, that’s a great question. When you think about the spike proteins and kind of where we are with COVID, it’s been a mad dash for everybody to generate the proteins and get them out commercially available. What we’ve seen, and this really holds true for any binning assay, but I guess it gets highlighted in the COVID scenario is that the protein quality can vary highly between these manufacturers. Tom, in his examples is using the ACROBiosystems’ spike protein for example. I know ACROBiosystems they do SEC-MALS for example, they check purity, and that’s critical for these binning assays to make sure that you have a really, really good quality protein to start with, because otherwise your binning data can be highly variable if you have heterogeneous mixtures of the antigen species, you’re binning against.
0:42:38.3 ND: That’s usually the biggest impediment to really making clearly discernible datasets. Somewhat secondary to that, obviously is the valency consideration, the spike protein as a trimer, you could also use a monomer form as well, depending on what your aims were, the trimer form would be best suited for a pre-mix type assay and then a monomeric form, a monovalent form would be best suited for a classical type assay, sandwich style assay. Probably the last thing just to note with these spike protein assays, and Tom was sort of hinting at that with some of the inhibition data they had, you can’t take that surface with antibodies arrayed and simply assess whether or not S1 for example and ACE2 can interact in the present being bound to these antibodies. So within the binning exercises, it’s a great way to demonstrate blocking profiles in addition to clustering the antibodies based on epitope.
0:43:38.1 JM: Thanks Noah. There’s another question for you. What is the minimal concentration of antibody needed for kinetic analysis on the LSA?
0:43:49.7 ND: I’d say it’s kind of variable, depends on the antigen you’re using and the surface, the antibodies, so we don’t wanna really say there’s a hard and fast rule, but we’ve seen successful enrichment of concentrations down… So the actual working assay concentration down into the mid-nanograms per mil range, so you can go quite low either, one you have very low expression or two, you may just have not much physical material and need to dilute it to sufficient volume for the assay itself, so you can be in the nanogram per mil range and run kinetics, it’s just sort of a lower limit of the assay.
0:44:28.1 JM: Thank you, Noah. Tom, there’s a question for you. Is Twist looking to develop these leads on your own, or are you looking to partner them?
0:44:38.8 TY: Like I mentioned before, we’re not a fully fledged drug development company, we are actually looking to partner with other zero and other companies and organizations to both assess the therapeutic values of these antibodies, so yes, we’re definitely looking for partnerships.
0:44:56.5 JM: Thanks Tom. Another question for you. How do you use next generation sequencing in your screening process, if you screen 3 by 384 and then transfer the hits to IgG.
0:45:17.2 TY: But we use NGS for two functions, one is to… Once the library is actually produced and put into the phage display vector we use NGS to verify that the output of all the sequences that we designed are actually in the final library and there aren’t any friendships or aberrant mutations. So it’s a QC step. The other step, the other way we use NGS is to track the enrichment of the antibodies as you go through the panning rounds, so as you go through rounds one, two, three, four of your phage panning, the idea is you’re enriching for binders, and what NGS allows us to do is actually quantitate which antibodies are actually being enriched over others. So that way, say if you have 300, 400, 500, 600 a lot of positive hits allows you to sort of prioritize which antibodies seem to be enriching in the panning itself by NGS. So it gives us some more data, and it gives us a way to triage if we need to triage.
0:46:26.0 JM: Here’s another question. It looks like you reformat your hits to IgG, and then do kinetics. Have you tried to include a tag in your library, for example, of these five tags, then you can use the LSA to directly screen your phage hits?
0:46:44.3 TY: So we have not, our typical… One of the reasons why we prefer reformatting the IgG is it just gives us a better… Well, first of all, once you’ve reformatted the IgG, it typically won’t develop these full IgG therapeutics, so we’re testing the final product or something close to the final product. And when you reformat a lot of these, you’ll see some drop out where they don’t retain binding once they’ve been reformatted from the SCC to an IgG. That’s an interesting concept though, it’s something that we might look into depending in terms of integration of the tag, they do… Our phage libraries do have HIS tags, but we can also just use other tags to couple, for the antibody fragments, rather than reformatting into the IgG.
0:47:48.6 JM: Tom, have you tested these antibodies for binding to S-proteins from related viruses?
0:47:55.3 TY: Not yet. We actually… That’s one of the things we’re going to be doing next. So we’re going to be looking to see if… Look at the cross-reactivity between, of binding to SARS-CoV-2 S1, also SARS, the original SARS spike protein and also MERS. We’re also looking to see if bindings are retained for some of the SARS-CoV-2 S1 mutations that appear to be popping up now. So specificity testing is definitely next in the dock for us.
0:48:36.2 JM: Is there an indication for functional superiority for antibodies that bind to monomeric RBD versus trimeric RBD?
0:48:45.7 TY: We don’t have that data yet. It’s something we definitely wanna see if there’s a difference. In our case it mostly seems that everything that binds to the S1 RBD also… Everything that seems to bind to the S protein trimer also seems to bind to our, the S1 monomers as well.
0:49:13.5 JM: Was there an assay development done to determine which conditions worked best?
0:49:19.4 TY: If you’re talking about the LSA itself, in terms of the regeneration additions, yes we did do some regen scouting. Obviously, when you do the… So for example the competition experiment, you wanna make sure that you have sufficient S1 immobilized to the chip to be able to see Ace2 binding and also be able to premix enough antibodies such that you’re well and lower access to fully block that interaction. So a lot of the scouting was just scouting different concentrations, making sure we had enough material to get a proper RU shift but not so much that you’re not able to block the interaction with the competing antibody.
0:50:09.5 JM: How do you design your library based on the antibody sequences recovered from patients?
0:50:16.9 TY: So I think this is similar to another question we had previously just… I can’t get into specific details, but the core idea is that you’re taking some of the CDR sequences in this library and then shuffling or mixing them with some of the sequences that we have that we know are coming from others human antibodies. So that we’re creating a semi-synthetic library. So the CDR sequences are lifted from these convalescent donors.
0:50:48.3 JM: In general, what’s the normal signal over background ratio, in your final round of phage library panning, round four? Is this target dependent?
0:51:02.1 TY: So the signal to background ratio in ELISA is basically looking at binding of the phage with the antibody display binding to either S1 or Ace2 and then looking at whether that same phage display antibody binds to a background protein such as BSA, or any other protein, and that gets us… That’s how we calculate our fold over background. Typically minimum for that is, you know, three times over our backgrounds, but if you look at the data on the slides, there are many there, there are five times, 10 times, 20 times over backgrounds, so we knew that we had a lot of very high affinity finders just from the phage ELISAs. And obviously a phage ELISA is a pretty crude, they’re more of a yes/no assay but that level of fold over background just did give us a hint that we had good picomolar, nanomolar level boundaries findings which did play through once we converted them and tested them on the Carterra.
0:52:01.4 JM: In your antibody selections, you have a low number of washes, do these include an incubation step where the beads sit in buffer?
0:52:14.8 TY: Yes, so typically we’ll do these on the… Either by… So they’re all cycle phase all the antigens are automated and coupled to strips encoded bead. We’ve played around with the wash conditions in terms of the number of washes. We’ve kind of discovered that if we run this on the Kingfisher automation that it’s… There’s not a lot of carry-over compared to when you’re doing this by hand, so the number of washes looks relatively low like three washes, four washes, five washes, but this is equivalent to doing it by hand, something like three, six, eight, nine washes by hand. So it’s more dependent on obviously the format in which you’re doing the panning and also by whether you do it by hand or KingFisher. I mean the primary way we track this is by phage, phage titer, so if the phage titers are going up and they’re not too high then we know that we’re on the right track.
0:53:09.5 JM: Thanks Tom. Antibodies against ACE2 sound less appealing than those against spike RBD, what are the proposed advantages?
0:53:19.5 TY: Well one potential advantage is that you have less risk of viral escape in terms of mutagenic escape. I mean the virus is obviously mutating quite… The virus has a potential to mutate very rapidly and you know we have to be able to test for binding. Obviously for antibody therapeutics, one of the proposed methods is by assembling a cocktail of antibodies so that, you know, you’re binding the different epitopes on the S1 concurrently so that let’s say one part of the S1 protein mutates, you don’t lose that mutatization ability.
0:53:57.0 TY: If you have an Anti-ACE2 binning antibody, you obviously have to test to make sure that it doesn’t negatively affect the actual catalytic function of ACE2. But you’re less likely to have some sort of mutational event where you’re no longer able to inhibit that interaction.
0:54:20.7 JM: Thank you Tom. We have time for one last question. How do you assess the risk of blocking ACE2?
0:54:27.5 TY: How do we assess the risk of blocking ACE2? So we haven’t looked at that yet. Obviously, we have to be able to develop our… Do additional assays to determine if our antibodies affect ACE2 function. It does look like, just from the S1 RBD ACE2 interactions, that for just blocking that interaction is quite a while away from the catalytic site, so there’s a potential that we’re not affecting function, but obviously we have to test for this. And we’re looking for partners to be able to do this. We’re not currently set up for something like this assay in our own lab yet.
0:55:11.7 JM: Thank you so much Tom, that was a great presentation. For those of you whose questions we were not able to answer, reach out to, get those questions answered. And I wanna thank you again Tom, Noah for your help on this webinar. And I hope you all have a great day.
Presented by Tom Z. Yuan, PhD, Senior Scientist, Antibody Engineering, Twist Bioscience