• Rapidly characterize large panels of antibodies for kinetics, affinity, and epitope diversity
  • Identify species cross-reactivity and target variant cross-reactivity
  • Obtain high quality data directly from crude hybridoma supernatants
  • Work with minimal amounts of sample
  • Build your IP portfolio

This webinar discusses how scientists at Dragonfly Therapeutics have successfully identified lead candidates from the immunization of different species of mice, as well as through yeast display approaches.

Identifying antibodies with high affinity, specificity, and epitope diversity is critical to antibody discovery campaigns. These campaigns can generate hundreds to thousands of potential lead molecules and incomplete characterization at the screening stage can lead to the advancement of antibodies with liabilities or a lack of strong candidates.

Presented by Dan Fallon, Senior Research Associate, Dragonfly Therapeutics

0:00:00.8 John McKinley: Hello, and thank you for attending today's webinar. At the end of the webinar, we'll have a panel discussion with Dan Fallon and Noah Ditto, Carterra's technical product manager. During the webinar, please send your questions by private chat to me, John McKinley, the host. Carterra's new nickel Biosensor chip is available for shipment. The antibody discovery industry has been making great strides in the fight against COVID-19. For example, AbCellera and Eli Lilly developed a therapeutic antibody for the potential treatment of COVID-19 in record speed, taking less than three months to advance from initial screening to human clinical trials, which is amazing. And the Carterra LSA provided the kinetic and epitope analysis for these studies.

Dan Fallon, senior research associate at Dragonfly Therapeutics, present antibody discovery for Bispecific therapeutics

0:00:50.2 JM: Today, we're privileged to have Dan Fallon, senior research associate at Dragonfly Therapeutics, present on their advances in antibody discovery for Bispecific therapeutics. Dan has a Bachelor's Degree in Biochemistry and a Master's Degree in Microbiology from UMass Amherst. Dan has been working at Dragonfly Therapeutics for three years and is focused on the characterization of biologics. Dan, I'll hand it over to you now.

0:01:20.2 Dan Fallon: Thank you, John. Thank you everyone for joining me today. Today, I'll be discussing how we implement Carterra's LSA system at Dragonfly for antibody development and discovery for a Bispecific platform. So at Dragonfly, our technology revolves around the natural killer cell. And so we aim to harness the broad anti-tumor activity of natural killer cells and redirect it against tumor cells. And natural killer cells are uniquely well positioned to attack tumor cells, because they require a very specific signal of both activating receptors and a lack of inhibitory signals and they bridge the innate and the adaptive immune system together quite nicely. So upon a certain combination of receptors being ligated on the NK cell, and missing inhibitory signals, this NK cell packed with granules of these perforins and granzymes will directly kill tumor cells by releasing them into the lytic synapse and causing apoptosis of the cancer cells, but can also secrete these chemokines and cytokines, which are really important for ramping up the immune system and the immune response.

0:02:39.4 DF: And so, these chemokines can recruit other cells of the innate immune system, specifically dendritic cells, and then these dendritic cells will be activated and then can go on and recruit and activate cytotoxic T cells. And these cytotoxic T cells can then go and directly kill the cancer cell as well. And so by approaching this response with the NK cell, we get a very robust and a very specific early immune response against a tumor cell, which gives us a very nice therapeutic window. So Dragonfly's approach is to use a Bispecific antibody that essentially bridges the NK cell to the tumor cell. And so one arm of the Bispecific will bind the tumor antigen with high affinity. And then the other arm will signal a receptor on the NK cell to tell it to release these apoptotic proteins and to ramp up this overall immune response. And so through this Bispecific platform, we aim to stimulate new mechanisms for both direct and indirect tumor killing. We know that the platform, what we call triNKETS, work powerfully as simple agents, we've seen that both in a number of models in vitro and in vivo, but also can be powerfully augmented when combined with T cell based therapies.

0:04:06.0 DF: And since these NK cells require such a specific orchestra of activating inhibitory signals, it does a very, very good job of discerning healthy and cancerous cells. And so usually with such a robust immune response, you'll get some nonspecific killing of healthy cells. But the NK cells are uniquely good at singling out these infected cells. And so this is a schematic of a traditional Bispecific antibody. And so we have this anti-NK cell arm pretty well established. And then as part of our discovery campaigns, we look for indications that we think may respond well to this early NK cell driven immune response, and that may... That the population may be lacking a good therapeutic. And so, we've been using the LSA to discover a number of new anti-tumor associated antigen arms for our Bispecific platform. So this is kind of a typical criteria we would look for in our antibody discovery pipeline. We usually want it to be, or we always want it to be high affinity for the tumor antigen, usually single digit nanomolar or picomolar affinity for a monovalent interaction, or an off-rate that is slower than 5x10-4.

0:05:35.7 DF: We usually are looking for sino cross-reactivity, if we think we can get it. We want high affinity binding to cell express receptors, both human and sino. We want no cross-reactivity with the related family members. So we want this antibody to be highly specific for the TAA. We want broad epitope coverage in our discovery campaign, so we can understand how targeting different epitopes can strengthen or weaken the overall response. We want a clean PSR profile and a clean HIC profile. So we want it to be highly specific, again, for that TAA and we want there to be a very limited number of surface exposed hypertrophic patches. So we want it to be highly manufacturable. We want it to be very thermal stable, so a melting temp of the fab to be greater than 65 Celsius. And then we want no potential sequence liabilities in CDR. So no glycosylation motifs or oxidation, deamidation, isomerization motifs so we can manufacture it in a very reproducible manner. And so today I'll be talking about a discovery campaign for a target that I will call Human TAA-1 and some of the challenges we faced and how we implemented the LSA to address those challenges.

Cartoon representation of TAA-1

0:07:00.0 DF: So this is a cartoon representation of TAA-1, but this is actually kind of what it looks like and these are real challenges associated with discovering antibodies against it. And so Human TAA-1 has low sequence homology with Sino TAA-1. So that is an obvious challenge for getting cross-reactivity for Sino models. TAA-1 is heavily glycosylated, and TAA-1 belongs to a wide family of highly homologous proteins. So there's what I call TAA-2, three, and four, look quite similar. There's a number of similar domains that are found in TAA-1, and we want the antibody to only bind to unique regions within TAA-1. So this is a pretty specific profile we're looking for, and in these discovery campaigns we can get dozens to hundreds of antibodies, and so it's critical that we can get a really robust characterization of these antibodies early on in the screening funnel, so we don't start advancing one that may be cross-reactive with one of these family members or it can't discern differentially glycosylated TAA-1.

0:08:09.6 DF: So we needed a really high throughput instrument to achieve these goals, and that's where Carterra's LSA came in quite nicely. It was able to achieve screening hundreds of antibodies at once, whereas the AK really wasn't built for that early on screening funnel. And so Carterra assists on the LSA can look at up to 384 interactions per experiment. It can capture directly from hybridoma supe. It requires low antigen, even for these really high throughput experiments, and the experiments can be done in the same day. You can set up, run, and analyze really in the same day. And these are some of the core applications that the LSA offers, the traditional kinetics affinity, epitope binning, mapping and quantitation, and below are just some representative output from these applications. This is non-regenerative kinetic sensor grams, a heat map from a binning experiment, across an experiment, a network map. I'm not very familiar with the mapping functionality, and you can also do some titer analysis and quantization. But for Dragonfly's purposes, we primarily relied on it for kinetic and affinity screening as well as epitope binning. But in the near future, we're going to start looking at the titer analysis functionality of the instrument.

0:09:35.9 DF: And so the LSA is able to achieve this massive throughput in large part due to its 96-spot printhead. It's this pretty cool component within the instrument that essentially can miniaturize your traditional 96 well plate into a... It's got a microfluidic gel in there and print it onto a traditional carboxymethyl dextran chip or whatever they're offering, whatever you seek to capture on it, and it can do this four separate times. So here is one 96-spot capture. It can stagger at four separate times for 384 captures. In a traditional unidirectional flow SPR instrument it doesn't use the space nearly as efficiently nor does it use the capture molecule as efficiently. So a regular capture requires, I think, maybe 250 or 300 microliters of solution per spot, but then will return, I think, over 90% of it back to the plate. So you're really only consuming 20 or 30 microliters per experiment.

We primarily used the instrument for hybridoma screening

0:10:47.8 DF: And so we primarily used the instrument for hybridoma screening, and so we don't always have 384 antibodies to capture, but we typically have a few dozen, if not up to 100, and these hybridomas have really wide titer variability. And prior to the LSA trying to adjust this on the Biacore was just super laborious, and one hybridoma might be 1 microgram per ml, another one might be 50 microgram per ml, and then these hybridomas are polyclonal, and so your capture level isn't always representative of what you can expect for an Rmax. And so it's a very imperfect kind of development on our Biacore system, and the LSA gets around this really nicely, just through just sheer number. And so what it can do is, say you have 96 antibodies, you can print at four different capture densities in one experiment. So rather than sitting at the instrument and capturing a few densities and then trying to implement that into your experiment, you can combine that assay optimization and the actual run into your experiment. And so that saved us a ton of time.

0:12:12.2 DF: Similarly, it can use the single-flow cell to dock onto the printed surface and uses, again, very minimal protein. So if you were to probe one antibody versus a full 384 array, you would use the same amount of analyte just based on how the single-flow works. And so, for an experiment on the LSA, if you were to do two full dilutions and run a full kinetic experiment you really only need 600 microliters of your top concentration, and then you can just do a two-fold dilution throw because it only needs a 300 microliter slug of analyte to get full kinetics across the surface. Whereas, if we were doing our Biacore, time aside, running on the Biacore would be a... 10 different experiments due to plate limitations and what it can actually run in one experiment. But just from the analyte standpoint, based on our typical association rate and our flow rate, we would use up to 170 mls, depending on your settings it might be 100 mls or 200 mls, but it's on that order. And so just very quickly comparing them, their ores and mag are too different, and so for this specific campaign this would have been extremely expensive. To purchase all the recombinant protein, we would've had to have probably bring it in-house and try to express and purify here, but since we have the LSA we can get around that quite nicely.

0:13:48.6 DF: And since the LSA has been implemented, we've actually had a number of hybridoma campaigns initiated at Dragonfly. Six, I believe, since getting in. It's been a little over year. And across these six, we've yielded almost 1000 different hybridoma supernatants to be tested, and depending on the program, there might be multiple targets. Like this one is actually what will be... TAA-1, what I'm discussing today. We're looking at human and sino binding as well as all of those different family members. So you're not just looking for 304 interactions, you're really looking for 1800. And so you can't get that done on another instrument and that's the area I think the LSA's fit really nicely in for us. And so overall, across these six programs in the past year we've had a look at over 4000 interactions. And so tying it back to our discovery campaign I'm discussing today, this was one of our most intensive ones, just in number of targets we wanted to probe as well as the number of supes we got from it. And so we had two different strains of mice, we had our wild type, BALB/c mice. We've had a lot of success in, of getting antibodies out of, but we also in this one tried transgenic mice, which have a Rat Fc but fully human variable domains.

0:15:18.0 DF: And so that can nicely accelerate our development pipeline because we won't have to go through a full humanization effort to graft it onto a human framework and try to retain the high affinity. And so these were performed by our collaboratives at Green Mountain Antibodies who always do a really wonderful job, and we came up with what we thought was a reasonable immunization strategy of using cell expressed protein, recombinant folding protein, and truncated versions of the antigen to try to guide the immune response to regions that are specific to TAA-1 and that may increase our likelihood of getting sino cross-reactivity. And then the ones from the BALB/c would have to humanize and lose affinity in that process and would have to affinity mature as well, which we do in-house. And then, as a second approach to kinda hedge our discovery campaign, we also pursued a yeast campaign where we took the splenocyte from the transgenic mice with the human variable domains and constructed six immune libraries.

0:16:29.8 DF: And now this is our screening funnel for hybridoma supes, and this is a bit specific to TAA-1, but generally follow a similar pipeline and this is a simplified pipeline. There's usually a number of decisions made and sub-steps to each one, but overall this is kind of how we approach it. So you have... Our collaborators at Green Mountain Antibodies, they do all the immunization and cell fusions and create the hybridomas, as well as doing the ELISA against TAA-1, human, sino, and then we'll look at cross-reactivity against family members two, three, and four. And then they'll ship us down all the hybridomas. And prior to the LSA, we wouldn't be able to really do any meaningful kinetic and affinity and specificity screening at this stage. We would have to kinda defer to the cell binding data, but since we got that in-house we can now run this and isogenic and cancer cell line binding in parallel.

0:17:35.3 DF: And so we can look to see if our SPR binding data it correlates well with the cell binding data, and it really speeds up our subclone nomination and sequencing selection process. And so we had... I think the instrument was undergoing a repair at this point. Usually we would find the supes that have a reasonable affinity for human TAA-1, and then we would bin those with tool antibodies we have, where we know what domain they bind and with... Wanna understand what kind of epitope diversity we have from the active supes. Unfortunately, we weren't able to use the LSA for this instance. It lends itself quite well with the instrument, but here we use Biacore and looked at what kind of diversity we had from this initial stage, and then based on the cell binding and specificity and affinity and epitope diversity we'll sit down and select subclones... Or, excuse me, clones for subclone nomination and sequencing.

0:18:42.4 DF: We'll get those back, hold tests, make sure the subclones retain their activity, and we will get the sequences, and then we will take those anti-TAA sequences and then plug them into our Bispecific or TriNKET platform, check what kind of diversity we have among those against each other, as well as see how they can kill in our cell-killing assays, and then we can look how the cell killing is correlated with the bin, how it's correlated with the affinity, and it's really a pretty complete package. Here is just a representative of what we would get from one of these preliminary kinetic screens. So very early on the pipeline, Green Mountain will send us down X number of antibodies, and that same day we can plate it and dilute it and run a buffer, and then just get it right on the instrument, and then that evening or the following day we'll have a pretty diverse panel, and an accurate panel of kinetics and affinities of these supes. And then this is, I think, a nice example of why that massive capture surface is actually a really, really nice feature because there's some people who were saying, "We don't need 384, that's overkill," but we found that we can actually really leverage that nicely, and that...

0:20:07.1 DF: Typically, if I have like, let's say I have 96 antibodies, I'll dilute, one to three of the supe to a running buffer, and then I'll do three full dilutions through there. And so these higher tighter clones like let's say, clone two here, all Rmaxes are pretty reasonable, and so all three dilutions work fine, but as you get to these lower tighter antibodies like clone three and clone four, the one to 27 isn't... Might not be the perfect dilution factor for that one, and then it's definitely not for clone four because there's just really not enough protein on the chip, and if that thing was polyclonal and you thought you captured something, but then you see no binding, well then you might be out of luck. You might just say, "Oh yeah, clone four is inactive," but in reality, you simply didn't have enough on the chip. And so we don't have to use these, we can just toss them out and say, "Here are three good representative sensorgrams of these clones and I picked these three because their Rmaxes were all pretty similar.

0:21:05.8 DF: And we can take all that data, we can triage out the ones that have too high of Rmaxes, ones that are too low, and if they're pretty similar, we can get some statistics on them. But once we look at the panel, we see we have a pretty diverse group of binders here with a fast off rate, slow off rate, and then when we compare the negative control to our non-binders, they look quite similar, so... The instrument does a very good job of really isolating the interactions at each spot, we don't see like bleed across if there was some interaction, these spots were near each other. It isolates them quite well. And so the ones that don't bind look just like our negative control. A really nice feature of the kinetic analysis software is this ISO-affinity plot and so if you want to understand your kinetic diversity of the library you get back. It's very easy to look at your affinity diversity, you can just sort, but this is a nice visual representation of what binders you have, so it'll plot the on-rate by the off-rate and the affinities are dropped across the plot and then you can go ahead and gate them.

0:22:24.7 DF: And so I gated this white region here with the off rate that's slower than 1x10 to the minus four and affinity that's tighter than 10 nanomolar and so we can go in here and say, "These are probably clones we're pretty interested in." And then there's these kind of two sub-gates where one is the high affinity, but they off rate might be a little bit faster, so probably ones we don't want to throw up but wanna take a closer look at, so the on rate looks like and the overall fit. Similarly these ones have a slow off rate but a weaker affinity, so it would be double digit plus a nanomolar and so again, we can very easily kind of triage this data.

0:23:04.3 DF: And another nice features that you can go ahead and just hover over each one, simply hovering over this one or one of these over here and the sensorgram will pop up with the kinetics and affinity, until you make calls to make sure you click this clone, but maybe it's a bad fit, but you can very easily investigate that. So that was the kinetic and affinity screening for the target, but we also wanna look at the other targets, the other family members, so here's a cartoon representation of TA1 and then TA2, 3 and 4 are all kind of just sub-domains of this but with slightly different glycosylation. And so pre-LSA, we just wouldn't have the throughput to look at 304 antibodies against each one of these, we would just have to defer to cell binding, if they were able to gather all that data or just wait until further in the pipeline where we could reasonably, actually check these on the Biacore, and usually we'd wanna do that with a purifying antibody or something but...

0:24:09.7 DF: The LSA does a nice job we just couldn't do a one shot concentration, I think with the one micromolar here, you can do a few here its a pretty quick experiment. And you can very quickly look at binders versus non-binders, and then throw them out, we can look at clone 43 has a high affinity for TA2 same thing for clone 19. Those maybe ones that were not interested in and can very quickly be triaged out and the experiments themselves are very quick, so if you were to do capture and run this binding experiment, it might take you like all of two or three hours, so you could do almost a full characterization of these three family members in a matter of a day to two days. Once we have characterized all the hybridoma supernatants, we will combine our SPR data with our cell binding data as well as the epitope tool map binning, we're not shown here, but it's another consideration. We'll look to see that they are in agreement 'cause we wanna make sure we're not advancing ones that look like they're high affinity to their common protein, but then show no binding to cell expressed, and so that's another check box we have to go through. And the LSA will actually improve our confidence in our recombinant protein because we can check hundred antibodies, and for the most part at least for this program our recombinant proteins match quite nicely with the cell express and so...

We see nice binding in the human TA1

0:25:38.3 DF: We see nice binding in the human TA1, we see a nice shift in the flow plot and then no binding to the other, the sino and the three family members that are plotted here. So those are in agreement and then here's just a truncated version of the human antigen to have high affinity, and then no binding. Unfortunately it wasn't cross-drafted to sino we found that that was a bigger challenge than we thought, and we were able to still advance the program without as many sino-cross reactive clones as we'd like. The hybridoma discovery screening funnel up unto this point resulted in a lot of really interesting well-characterized and well behaved antibodies in our hands, from 306, we've whittled it down to 47 binding the recombinant protein, the recombinant human. And then 10 binding the sino, then we found 26 binding the isogenic cell expressed TA1 and then 17 through this cancer cell line, and then 11 out of the group show no cross-reactivity to the family members both by SPR and by facts.

0:26:48.1 DF: So this was us kind of summarizing their clones we thought were interesting. From the Hybridoma Supernatant stage, and so for most part they are single-digit nanomolar affinity. A few of them are double-digit, this one's a bit weaker, and I believe this was selected because it had a really unique epitope. And so, we'll take a number of those just to have a better understanding of how the epitope, what role it plays in the killing. And so this was from the BALB/c mouse. And then what was interesting, these transgenic mice had a really, really high affinity, somewhere in almost a 100, 200 Picomolar range affinity for human TA1, whereas some of them didn't show any binding, but I believe showed some binding to cells and had a unique epitope.

0:27:35.1 DF: And so, the recombinant proteins aren't always exactly representative of the cell-expressed but for the majority of this campaign, we saw a good agreement across it. So we'll select those clones to be sub-cloned, we'll recheck their activity to make sure we didn't lose anything in the sub-cloning process. We'll gather the sequences, and then we will plug it into our bio-specific platform. And we'll check cell killing but we can also very nicely check the cross-binning because these aren't from supes anymore, so we can directly mobilize them. And what we found on the preliminary binning experiment was that there was, there are four tool maps, and the majority of them didn't bind to any of them.

0:28:25.6 DF: And so, we selected the ones that showed unique profiles but different sandwiching antibodies. But the majority of them didn't show, bind to all four. And so, it looks like we have some epitope diversity here, but it was somewhat biased to one region as there's these communities here, but there's a clear overlap between them. And so, we had epitope diversity, but it just looks like it was biased towards one, possibly one domain on that antigen. And then we can take that cross-binning data and compare it to our cell killing data. And so we'll incubate NK cells with the cancer cells, and our bio-specific, and we'll check to see how well those bio-specifics can bind the cancer cell line and can signal to the NK cell to kill it. And in this specific one we found, the majority of them were already biased to this domain, so it was difficult to tease out whether the domain had a big role, but we found that, overall, the affinity had a significant role and it's both a TC50 and it's a specific lysis.

AB-073 is its highest percent specific lysis and lower CC50

0:29:37.9 DF: So AB-073 is its highest percent specific lysis and lower CC50 had an affinity about nine nanomolar. And similarly, these other two, AB-150 and 52, had affinities on the same order of nine nanomolar. And then these ones drop up pretty quickly, and that's more getting to a 20 and 25 nanomolar affinity range. And so, this helps us better understand the profile we're looking for in these antibodies and whether we need to go back and if they mature or potentially look at a different domain, but in this instance, it looks like the response was a bit biased to what it reacted against in the mice. So overall, the LSA has allowed us to just do much more rigorous screening, much earlier on. So pre-LSA we weren't screening cross-reactivity against recombinant proteins. If we had a prohibitively large library and if we had 10 or 12, we can still get it on in a reasonable time, but once you get dozens to 100, it just makes no sense, and so we couldn't do it.

0:30:49.1 DF: We'd have to defer to the cell-binding data and then, again, and work through the pipeline until that number shrunk to something that's more manageable. But now we can get a much more robust package early on, and so we don't advance an antibody that is potentially cross-reactivity. That has cross-reactivity to a family member that we don't want. Conversely, we don't want to triage one out, based on incomplete characterization. So we can avoid that problem pretty nicely with this instrument. And so, majority of it so far has been kinetic screening, and some epitope binning and cross-binning experiments. The future applications we're gonna look at are the titer analysis of the anti-bodies because we'd like to provide that information to our colleagues in doing this cell binning, and there are other in vitro assays. We're also look at Supernatant cross-binning. I've had a few conversations with Ira at Carterra, about the possibility of very early on in the screening funnel, taking those crude hybridomas, capturing them to the surface, and then cross-linking them so they covalently link to the chip surface.

0:32:06.5 DF: And then binning them against each other, and that we can have a much earlier understanding of what kind of diversity we have, rather than having to do it for just to tool antibodies. And with that, I'd like to thank a few folks at Carterra who made this all possible, John McKinley spent a long time organizing these webinars, and going back and forth, and making sure they're organized, and just assembling, really, the whole presentation, similarly, Noah Ditto had a large role in facilitating all of this. Ira, back at Carterra and he's their application scientists, has spent tons and tons of time with me, going over data analysis, experimental design, and so a lot of the data you saw today was birthed out of Ira's help.

0:32:55.7 DF: And Ric Galle is their Senior Field Service Engineer, who if we ever have an issue with the instrument, he pops in the next day and has it up and running again, so he's been a tremendous help. At Dragonfly, Souvik Chattopadhyay who is our Cell Expression Head, in our biologics group. He is also the project lead of TA1, and so he spent a long time with me just going through the screen funnel because I didn't at the time, I didn't have all the information on all the cell-binning and cell-killing, and so he helped me very nicely assemble some really critical pieces of data as well as just going through the full screening funnel. And lastly, Asya Grinberg is our head of biologics at Dragon Fly, and she took a lot of time to review the slides and help.

0:33:44.3 DF: Make a more compelling story about how we implemented it, and some of the important features to point out, and so she's spent a long time with me just going through what's important, what's not, and helping me to complete the slide deck for today, and with that, I think Noah is on the line from Carterra and he can probably answer some of the more specific questions about the ins and outs of the instruments, and I'd be happy to take any questions about how we use it at Dragonfly or anything that I can answer, Dragonfly related.

0:34:19.1 JM: Thank you, Dan. That was a very informative presentation, we'll start our panel discussion now with Dan and Noah Ditto, our technical product manager, if you have any questions, please send them to me, John McKinney the host by private chat. So we have some questions coming in. Dan, do you generally see agreement between recombinant and cell expressed antigen binding? 

0:34:49.8 DF: Yeah, we usually do pretty rigorous characterization of the recombinant protein before doing any screening, and so I think that's a function of the biologics teams just doing a deep characterization of them, and so that's always an important part of our screening funnels to make sure we have good recombinant protein, otherwise, we wouldn't be producing very useful data. So for that fact, yeah, usually we have a pretty strong agreement between the cell expressed and recombinant protein.

0:35:23.0 JM: Another question for you, Dan, do you use the LSA to assay TriNKET molecules or all of those qualification tests done with cell based assays? 

0:35:33.6 DF: No, we definitely use the LSA for that, especially if we have to humanize a number of the clones from the wild type BALB/c mice, and so if we have a lot of humanized variants, we may go to the LSA to check that we didn't lose binding, or how the affinity and kinetics change of that.

0:36:02.3 JM: Okay, we have a question for Noah. How long does it take to complete kinetic screen for 384 clones? 

0:36:18.5 Noah Ditto: It depends a little bit on how you're setting up the assay, but there is some upfront surface prep considerations you might be making it anti Fc capture surface or already have one made, and it kind of depends on how many blocks of 96 unique samples, you're printing on the surface 'cause it can go from 96 all the way up to 384 to use the maximum capacity of the sensor chip in the LSA, but maybe leaving those variables out the actual kinetic analysis itself, which includes warm-up cycles of buffer blanks and then a kinetic titration of maybe an eight-point member series would be on the order of maybe six hours, give or take, depending on your off rates, if you have really long off rates it may go a little longer, if you're doing a little more of a rapid characterization, but shorter off rate times it might be a little less.

0:37:15.9 JM: Another question for you, Noah, how reproducible is lag and capture between cycles and between spots? 

0:37:24.2 ND: It's usually fairly good. It's, again, kind of a qualified answer I'll give knowing that there's lots of different systems and ways you can run assays, so there's certainties and variables involved, but if we just look at maybe a fairly standard anti Fc sort of capture approach, something like that. We would expect to see probably a 10% or less variability typically in leading densities across the array, usually, that's kind of our benchmark in manufacturing to really get under that 10% to make sure the instruments perform that good. And in the field, we typically see it perform that way as well.

0:38:03.0 JM: Another question, Noah. How good is bulk shift referencing for low affinity experiments? 

0:38:15.2 ND: So I am not entirely sure maybe what the concern is with low affinity bulk shift but I'll address bulk shift in general. So if we're doing crude experiments, there could be bulk shift being introduced from the crude matrices themselves, software does have the ability to do double referencing to remove any of those artifacts, so it is not an issue. Sometimes in binning experiments, we may have the Supernatant flowing and have a fairly large bulk shift for example, that we just adjust the data analysis parameters to avoid the influence of that bulk shift on the data for example, but usually... I guess the question is maybe affinity, if you're flowing a low concentration material and you might need more material we typically run the assay in other ways with crude supernatants for kinetics with the antibodies on the surface kinda like Dan was describing here. So generally, we purify basically on our chip when we array the antibodies on the surface, so it tends to mean that we don't have any crude matrices. The antigens typically purified. So there's not usually a bulk shift concern in terms of trying to measure kinetics.

0:39:36.6 JM: Thanks, Noah. Dan, how long does it take for you to set up and run a kinetics experiment and also for a binning experiment? 

0:39:48.2 DF: So Kinetic, we're doing it from hybridoma supe, so really the most laborious part is just taking supe out of tubes and putting it in the plate. So to get through like if you wanted to plate out 96 supes, it might take only a little under an hour, and then the navigator software is very easy to set up a run, so actually putting it on the instrument takes all of five minutes, the same thing with diluting the antigen. So the upfront work is really just setting up the plate. And then...

0:40:20.9 DF: And then diluting the antigen and writing the method takes again, like five minutes, so I'll say about an hour. But then for binning, that can be more involved depending on what kind of experiment you're doing. So if you are just, if you're doing a large cross-binning experiment like that one I had showed with the purified material, that can take a little bit just to create different chip densities of each immobilized antibody, as well as to then go and dilute it in a running buffer and come up with a consistent concentration of that. So that can vary a little bit. It could be half an hour give or take. That one part took me two hours to set up, but there's really no, I think, alternative to getting that kind of breadth of data.

0:41:15.7 JM: Thank you, Dan. There's a follow-up question about the time it takes to do the analysis for a large kinetic experiment, say, over 100 clones. And also about doing a binning analysis for a large set of clones for the analysis.

0:41:32.0 DF: Yes, so Kinetics is nice because the software offers a lot of tools to just screen out dead antibodies or ones beneath certain Rmax thresholds. And so the more I use it, the more I appreciate those kind of functions. And so if I can go through like 96, I take a look at every sensorgram, just to see that one analyte concentration didn't have some bulk shift or something, or the supe was polyclonal, but usually we'll have anywhere from 10% to 50% be active, and so for 100 antibodies, it might take me maybe an hour to go through them. Binning takes longer because there's just a lot more sensorgrams and more, I think maybe not entirely subjectivity, but I think more detail analysis, cycle by cycle you have to take in different considerations, so that one will take a bit longer depending on, if you wanna really get that exact measurement. It would, it might take two hours or three if you wanna go sensorgram by sensorgram, like sandwiching event by sandwiching event.

0:42:49.2 JM: Thanks, Dan. Noah, if you have lower expressing supernatants, what is the minimal concentration required for kinetic analysis? 

0:43:03.7 ND: Again, another question, which is assay dependent on what your sample types are, but we can say that in anti-FC type capture format, we really leverage the power of the LSA to have its bi-directional flow capabilities. So when we take crude matrices in particular, which expression levels can be low in variable, that sample plug is flown back and forth across the surface for really any amount of contact time the user wants to. So we're doing this on-chip purification or enrichment, it really drives the assay sensitivity to really, really low levels. So data sets we've seen for anti-FC capture group, we'd be pretty confident that you can get in the maybe 50 to 75 ng/mL range, and that's getting enough material of say the antibody on the surface that you can make a confident kinetic measurement with a typical protein antigen. So that would be the ballpark, 50 to 75 ng/mL.

0:44:12.5 JM: Noah, another question for you. Can I set up multiple analytes to run over the same sample set for assays one after another? 

0:44:23.3 ND: Yeah, and it would sort of depend on how you're setting up the array and what the reactivity profile is of the antibodies in the array. So if you happen to be doing a multi-flex experiment, maybe you have two different projects of smaller sets of antibodies and you simply wanna run a single experiment, you could capture, say, antibodies from Project A and antibodies from Project B onto the same sensor chip. And then in one experiment flow antigen against Project A and antigen against Project B across that surface, you would get two measurements and they would be, if it goes well you should have no cross-reactivity in your antibodies. So you can generate the data that way. Conversely, if you directly immobilize the antibodies and you have maybe a human and a Sino form, you would have the ability to maybe run the human series regenerate the surface of any human antigen, and then run the Sino series, for example. So there's a couple different ways to do it, and it really just kinda depends on what the antibodies are and what they're reacting against. It dictates the assay set up.

0:45:33.7 JM: We have one more question. Dan, how were you screening hybridoma supernatants before you had the LSA? 

0:45:43.5 DF: Mostly, mostly through cell binding. We would test them and measure kinetics and affinity, after the Biology team would let us know which ones were active and which ones bind the cell express version, but the screen funnel from the SPR end was, I guess, more primitive in that we just, essentially we didn't really have much for the really large libraries, again, if we got them, the smaller ones, piecemeal, we would still be able to work through it reasonably on the Biacore, but as we kind of increase these libraries, that was such that the funnel would be deferred to cell binding and then see what was left, and then we would get those on the Biacore.

0:46:39.7 JM: Well, great. Well, thank you so much, Dan, for that very informative presentation. Thank you, Noah, for joining our panel discussion. And I would like to thank our audience and everybody for attending today's webinar. If you have any additional questions, please reach out to Dan and Noah, their email addresses are up on the screen, and we really appreciate everybody attending. Thank you all so much.