PD-L1 is a key inhibitor of T cell activation that is often over-expressed in cancer to escape immune surveillance and promote tumor progression. Blocking antibodies against PD-L1 or its receptor, PD-1, have shown significant clinical benefit in some patients with PD-L1 expressing tumors. Hence, there is great interest in generating therapeutic antibodies against these targets to counteract the immune suppression mechanism that tumors rely on for survival.
Epitope binning studies on the Carterra LSA revealed that hybridoma antibodies covered more diverse epitope regions but potency in functional assays was weaker across all bins while B Cell Cloning antibodies were concentrated in fewer epitope bins that represented greater potency.
Our data show that plasma B cells secreting functional antibody candidates can be rapidly identified on the Beacon compared to several months for a hybridoma campaign and screening those antibodies for epitope specificity and affinity can be completed in just days on the Carterra LSA, thus substantially accelerating the antibody discovery process.
Presented by Shireen Khan, PhD, Senior Director of Biologics, ChemPartner
0:00:00.4 John: Today, we have an exciting presentation by Shireen Khan at ChemPartner on the discovery and characterization of potent anti-PDL-1 antibodies in just weeks. Shireen is the Senior Director of biologics at ChemPartner, where she leads a group that has expanded ChemPartner’s capabilities into single B cell cloning on the Beacon platform. She also leads multiple therapeutic antibody discovery programs for biotech and pharmaceutical companies. Prior to ChemPartner, Shireen led the functional biology group at XOMA, and advanced several therapeutic candidates through in vitro and in vivo efficacy studies. Shireen received her PhD in Biology at the University of California, San Diego. After today’s presentation, Anupam Singhal from Berkeley Lights, their Product Manager for antibody therapeutics, and Noah Ditto, Carterra’s technical product manager will join us for a panel discussion. Please submit your questions with the question box on the right. I’ll hand it over to you now, Shireen.
0:01:12.1 Shireen Khan: Thank you, John, and thank you to Carterra for inviting us to present our proof of concept study that we did with Berkeley Lights. So we did a project with Berkeley Lights to discover potent anti-PDL-1 antibodies on the Beacon platform by single plasma B cell cloning. Once we selected the candidates and expressed and purified them, we were able to collaborate with Carterra on the LSA platform to do some epitope binning analyses as well as evaluation of all of our candidates that came from the single plasma B cell cloning approach versus a parallel Hybridoma approach that we had done in our facility in Shanghai. So by being able to evaluate all of the candidates side by side, we really had a good idea of what the relative affinities of those antibody candidates were, and also the relationships between the antibodies with respect to the epitopes.
0:02:23.0 SK: So in terms of antibody discovery approaches, we have a few to select from, including phage display, Hybridoma and B cell cloning, so each of these methods has their pros and cons, in particular for duration phage display, as well as B cell cloning can be a quicker approach to identifying the best candidates. In terms of the repertoire for phage display it’s either natural or designed, and by both Hybridoma and B cell cloning, it’s a natural repertoire because typically you’re relying on animal immunizations, but the number of members that you can evaluate by B cell cloning tends to be far greater, certainly on the Beacon, it’s in the tens of thousands of single B cells that we can evaluate, compared to around 10,000 by Hybridoma technology.
0:03:18.1 SK: For each of these methods, again, there’s pros and cons in terms of antigen, so by phage display, there’s more flexibility in terms of the types of antigens that you can work with, but it tends to be more challenging for targets where there’s limited extracellular domains such as GPCRs or ion channels, but we’ve had success by animal immunization with either cells over-expressing the target or genetic immunization by Hybridoma or B cell cloning, so accessing these more challenging targets is feasible by those two methods. In terms of affinity, there may be a need for affinity maturation from a phage display library, whereas for both Hybridoma and B cell cloning, you’re taking advantage of that in vivo affinity maturation that occurs and the screening format, of course, varies between phage display being scFv or Fab versus Hybridoma, and B cell cloning is directly looking at the IgGs.
0:04:20.9 SK: So all of these methods taken together, we really see B cell cloning as an opportunity to take the best of both worlds that come from either phage display or Hybridoma, so for B cell cloning, again, no need necessarily for affinity maturation, if you can pull out good candidates, based on that, in vivo affinity maturation, there’s potential for greater sampling of the immune repertoire than you can do by Hybridoma, accelerated discovery similar to phage display, but you also get the sequences early similar to phage display, but what we really think is the advantage of B cell cloning is the opportunity to bring function forward in the drug discovery process and be able to identify candidates earlier in the process that have the ideal functional properties.
0:05:20.1 SK: So at ChemPartner, we have the Berkeley Lights Beacon platform in order to accelerate our antibody discovery process, so on the left there is a picture of the Beacon platform, it can accommodate up to four of the OptoSelect chips as shown in the middle of the slide here, and on these chips, there are thousands of these sub-nanoliter chambers called NanoPens, and on the platform, we’re able to use controlled light patterns to move single B cells around and sequester them within these NanoPens, and by doing so, we can allow them to culture and secrete antibodies and subsequently assay for target specific binding as well as functional properties.
0:06:10.9 SK: So what we see as the value proposition for this platform is being able to do functional assays upfront. So on the left there, you can see that there’s beads and then potentially cells that can be loaded into the channels of these chips. And so if you have your beads coded with the target of interest or cells over-expressing the target, it’s possible to see blooms above the pens where there’s beep B cells secreting target specific antibodies, and you see that nice accumulation of signal right around the NanoPens where those B cells are located. And so it’s possible to do multiplexing assays in this format to look for antigen specificity as well as cross-reactive binding. So in the past, we’ve been able to do in a multiplex format, binding to human, mouse as well as proteins.
0:07:08.6 SK: And so other potential assays that we could develop on the prep platform include cytokine release, potentially epitope binnings, experiments can be done right on the platform as well as gene expression assays and cytotoxicity assays, so we’re constantly thinking of new assays that we can develop for the platform, really enable this capability. So this is what the B cell cloning antibody discovery workflow looks like. Of course, starting from the immunizations, we will harvest either the bone marrow lymph nodes or spleen from the immunized animals and then do the isolations in our South San Francisco location, we’ll do the loading onto the Beacon as well as binding and functional assays, export the hits and prepare the CDNA, do the VH/VL amplification, we send it out for sequencing, and then once we get those sequences back, we will do the cloning expression and purification in our Shanghai facility and potentially functional characterization of the purified antibodies.
0:08:26.3 SK: So we initiated a collaboration project between ChemPartner and Berkeley Lights in order to generate antibodies targeting PD-L1, and there are several clinical candidates that we could actually directly compare the output from B cell cloning compared to a standard Hybridoma approach, which is what we would typically do for our campaigns such as this. So we wanted to basically evaluate the functional properties of the antibodies, including binding, affinity, diversity of sequences, as well as epitope coverage. We had used identical reagents for immunization by IP and similar assays were done at the screening stage, either on the Beacon or by hybridoma supernatant screening.
0:09:19.4 SK: We harvested the spleen and bone marrow for both of these methods, and we used two strains of mice for the hybridoma campaign, we standardly use SJL as well as biopsy. And on the Beacon platform, we only used biopsy animals for this study. The detection methods were quite different, so on Beacon it’s an image-based detection system, whereas by hybridoma supernatant, we were either doing facts-based finding or ELISA binding and blocking assays. And so ChemPartner was set up uniquely to be able to provide all of the reagents needed for this proof of concept study. So we generated the reagents, the recombinant proteins for immunization, the cell lines over-expressing either human or sino PD-L1 as well as generating the recombinant antibodies that we then characterized in functional assays. So the Shanghai team developed binding assays as well as T cell activation and mixed lymphocyte reaction assays so that we could evaluate the functional properties of our hits from each of these methods.
0:10:45.0 SK: We did the hybridoma approach in Shanghai, of course, and then the B cell cloning was done in collaboration with Berkeley Lights. We did the VH/VL sequencing in Shanghai and then characterized the purified antibodies through a series of functional assays.
0:11:09.7 SK: Okay, so just comparing the B cell cloning compared to hybridoma. If you look at the amount of time it takes to get to the sequences, now, this is not including the immunization strategy and the amount of time it takes to do that schedule, but for B cell cloning, you can pretty much get all the way through the VH/VL sequencing of your hits within two to four weeks, whereas by hybridoma, after going through the entire process of fusion, ELISA binding as well as cell-based binding and blocking assays in the primary and secondary screening format, followed by selection of your high priority candidates for production and purification, you actually have about three to five months until you get to the sequencing results.
0:12:06.0 SK: But the advantage here of course, is that you’re characterizing purified antibodies before you’re selecting the candidates for sequencing. Unfortunately, the downside of that is that sometimes it’s very useful to have the sequence information upfront so that you can select candidates based on their diversity of sequence and potentially evaluate whether you should select a set of leads and back-ups based on the sequence diversity. You can also look at liabilities in the sequence, but without that information and going through the entire process in selecting the leads can be a little bit risky.
0:12:53.6 SK: So the method that we used to evaluate blocking properties by hybridoma was basically an ELISA-based selection method. So we did a standard blocking or non-blocking assay by ELISA, looking at whether biotinylated PD1 bound to the human PD-L1 ECD-Fc in the presence of the anti-PD-L1 antibodies. And so in order to identify and select the hits, we cast a broad net, so the blocking activity that was anywhere in the range of 30-100% was selected as a hit at the primary screening stage using hybridoma supernatant. So it’s a quantitative method, but casting a broader net would enable us to pull more blocking antibodies forward and then mitigate the risk of the potential for those clones dropping out during the sub-cloning process.
0:14:06.1 SK: And so the cell-based binding and blocking assays on the Beacon platform are image-based, so it’s quite a bit different from that perspective compared to ELISA or facts-based methods. But basically what we did for this particular project was load the pens with CHO-K1 over-expressing human PD-L1, and in that same pen is actually a B cell secreting target-specific antibody. We also imported beads that are coated with human PD-L1. And we assayed for in-channel bead binding as well as in-pen cell-based binding. And so in order to evaluate the blocking versus non-blocking activity, what we did then was to import soluble PD-1 that was fluorescently labeled.
0:15:03.7 SK: So basically, if the PD-1 was able to bind, that would indicate that it’s a non-blocking antibody, as shown on the bottom right panels. And if it was able to exclude… If the antibody was able to exclude the ability of PD-1 to bind, you’d see very illuminant signal in the FITC channel, as shown on the upper right there, and you would see a strong signal in the Texas Red channel, suggesting that the antibody is present and it’s binding to either beads or cells. And so that’s how we categorized our antibodies in terms of blocking versus non-blocking activity, and so it’s just a very different method, but we looked for antibodies that had that stark contrast in terms of signal for the PD-1 in the Alexa Fluor 488 channel.
0:16:01.3 SK: This slide shows the outcome of the B cell cloning screening on the Beacon platform versus the hybridoma primary and secondary screening results. On the Beacon platform, we evaluated over 33,000 single B cells, and we looked at both bone marrow and spleen, and found that there were over 200 cell-based binders to human PD-L1, and a total of 35 blockers were identified. If you look at the distribution between the number of blockers from the bone marrow versus the spleen in this experiment, it look like most of the hits came from the bone marrow. We’ve done subsequent experiments and found that those numbers are usually more even between spleen and bone marrow, but it was interesting that a lot of these hits actually came from the bone marrow in this particular experiment. And in the hybridoma campaign, we found that a majority of the hits came from the spleen, in fact, and that’s not surprising given that there was limited success with the fusions that came from the bone marrow.
0:17:14.0 SK: So basically, we got more binders and blockers identified through the hybridoma campaign because of that broader net that we cast for identifying or selecting a candidate as a hit. And so we had to narrow that down at the secondary screening stage to around 50, that we would then move forward into the sub-cloning stage. So you can see the total numbers of hybridoma clones screened are a little over 13,000, and 48 blockers were pulled through, compared to the 35 blockers from the B cell cloning method.
0:18:02.1 SK: So the next step in the process was to do the single B cell sequencing. So out of the 35 blocking hits that we had, we were able to recover good sequences for 22 out of the 35, and so that’s a recovery rate of about 60… Between 60% and 70% in this particular experiment, but since then we’ve been able to significantly improve that recovery rate, and it’s more around the range of 80% more recently. Those were then expressed and purified in Shanghai, and at the same time, 40 of the hybridoma antibodies that had the desired binding and blocking activity were purified, and so we compared the activity of each of these antibodies across multiple assays. So we looked at binding, as well as blocking, affinity, and then the epitope binning profiles. And then we did additional functional assays, including T cell activation compared to the benchmark antibodies, which are shown in the table below here, several anti-PD-L1 antibodies from BMS, Roche, as well as MedImmune/AstraZeneca.
0:19:23.0 SK: So this is an example of the binding properties of the hybridoma versus B cell cloning antibodies, looking at binding to CHO-K1 over-expressing human PD-L1. On the right is the benchmark antibodies, and you can see for the hybridoma antibodies in general, we saw a broader range of maximum MFI by FACS compared to the antibodies coming from B cell cloning. So it looked as if there were a lot of antibodies that had very similar profiles by B cell cloning, at least in this campaign.
0:20:05.8 SK: In this slide here, we’re looking at the binding profiles of hybridoma versus B cell cloning antibodies to the cells over-expressing Cyno PD-L1. And again, in this example, we’re seeing that the maximum MFI varied for both the hybridoma as well as the B cell cloning antibodies. And recall that we did not use Cyno PD-L1 binding as a criteria for selecting our hits on the Beacon, so from that perspective, it’s not so surprising that some of these candidates don’t bind to Cyno as well as others, and so there’s a broader range of what we ended up with likely due to the lack of including this as a screen up front on the Beacon.
0:21:03.5 SK: In this slide, what we’re showing is the outcome of the blocking assay results, the PD1/PD-L1 blocking assay by ELISA using purified antibodies. And what you can see at the bottom there are the benchmark antibodies for both B cell cloning as well as the hybridoma experiments, which were done at a different time. And you could see, most of the time, the IC50 values across the three different benchmark antibodies are around 0.5 nanomolar.
0:21:37.1 SK: And so if you compare that to the results with the B-cell cloning antibodies or the hybridoma antibodies, what you can see is that the B cell cloning antibodies primarily had a tighter range of IC50 values, ranging there from about 0.4 to about 0.7. Compared to the hybridoma antibodies, there was a broader range of IC50 values, so that ranged from about 0.6 all the way up to 1.3. And again, this was likely very much due to the method by which we use to screen and select the blocking antibodies being very different on the Beacon as sort of an image-based selection criteria that was probably very stringent compared to the hybridoma in which we cast that broader net.
0:22:37.6 SK: So the next studies that we did included affinity estimations of our antibodies using the Biacore. This was done on a Biacore 8K, and what we did was captured through the Fc, the anti-PD-L1 antibodies, and then we flowed the human PD-L1 ECD-His material to be able to estimate the affinities.
0:23:10.4 SK: In this slide, we’re looking at the iso-affinity plots using the Biacore, and what we can see here in red are the benchmark antibodies, which are all sub-nanomolar affinity antibodies. And by the B cell cloning method, we were able to identify at least two antibodies that had affinity in the same range as the benchmark antibodies, and there were several other antibodies that had affinity and higher off-rates than the benchmark antibodies.
0:23:49.3 SK: On the right-hand side, you can see the sensorgrams for atezolizumab compared to P1A6 and P1A2, which were our two B cell cloning antibodies that had the highest affinity. And so what you can see is that at least two of the candidates that we identified had reasonably fast on-rate and slower dissociation curves and similar profiles in general compared to the benchmark antibody. And so, if you think about the number of B cells that we screened, 33,000 total, it’s obviously a very low frequency event to find antibodies that actually had the sub-nanomolar affinity, and so that kind of underscores the need to screen deeper into the immune repertoire to pull out candidates that actually have properties similar to this.
0:24:52.6 SK: So we were very excited about the prospect of evaluating our B cell cloning as well as hybridoma antibodies on the Carterra LSA. This platform itself was purpose-built by Carterra to support antibody characterization by conducting a lot of screening using minimal sample volume. So the LSA name comes from lodestar, a reference point used for navigation, and it kind of alludes to the purpose of the instrument, which is to help scientists take large pools of candidates and figure out the optimal candidates to move forward with based on biochemical properties.
0:25:42.3 SK: So the LSA is unique among biosensors based on its ability to switch fluidic modes between the arraying of up to 384 ligands, followed by single injection across this array in a single-channel mode. So from each injection, real-time binding signals are measured for the 384 active surfaces, plus 48 separate interspot references. Additionally, the bidirectional flow used in creating the array and associating analytes during binding analysis allows for robust signals to be measured out of very low concentration samples such as crude supernatants like you’d find in a hybridoma screening mode.
0:26:31.1 SK: So the applications that can be leveraged on the LSA platform include affinity measurement assays either by global kinetics or steady-state affinity. In addition to that, epitope characterization, such as epitope binning where antibody sharing common epitopes are clustered or epitope mapping which identifies domains or residue level binding sites. And lastly, it’s possible to determine the antibody concentration such as those found in crude supernatants, which is quickly done with the quantitation workflow, and this can be useful if it’s necessary to normalize the amount of protein present in crude supernatants.
0:27:19.8 SK: So, what we did on the LSA was to do epitope binning of the antibodies from B cell cloning versus hybridoma, so we had 24 antibodies from B cell cloning and 41 from the hybridoma campaign, and they were evaluated for kinetics as well as epitope binning. Covalent array was prepared using two antibody concentrations, and the array was first used for binding kinetics using four different concentrations of the PD-L1 ECD, and then the same array was then used to do a classical sandwich epitope binning assay in a pairwise manner.
0:28:00.9 SK: And so, the instrument ran for 30 hours, generating over 35,000 sensorgrams, and that would have taken about a month to complete by Biacore 8K. What we also found very useful in this analysis was to generate through the software on the LSA the community plots, so we were able to categorize our antibodies based on the relationships of the binding properties in the pairwise sandwich ELISA assay, or epitope binning assay.
0:28:44.0 SK: Shown on this slide is the iso-affinity plot that was generated on the LSA, looking at the B cell cloning in green, hybridoma in blue, and benchmark antibodies in red. And you can see there are several hybridoma antibodies, at least seven, were identified with B cell cloning and 15 were identified from hybridoma that had high affinity. If we compared the correlation of the LSA to the 8K, we found that there was really good correlation between these two methods with a fairly high R-squared value, so the data coming from two of these methods did line up very nicely.
0:29:37.1 SK: And in this analysis is the network plot. And so on the left-hand side, we are looking at the antibodies coming from either B cell cloning or hybridoma, so B cell cloning in red, hybridoma in blue, and the benchmark antibodies in yellow. And you can see that the number of bins that the blue colored antibodies are in is representative of 13 total bins, and so that’s broader epitope coverage from the hybridoma antibodies compared to B cell cloning, which actually clustered primarily in four different bins. And you can see that the B cell cloning antibodies also were categorized in the same bins as the benchmark antibodies in yellow.
0:30:27.8 SK: So if you toggle your eyes back and forth between the plot on the left, which is colored by antibody source, and the one on the right, which is colored by potency in the blocking assay, so the IC50 values, you can see the general trend that if the antibody is red on the left, it tends to be shades of red or darker orange on the right, meaning higher potency in the blocking assay. Whereas the antibodies that are blue tend to be varied in terms of the gradient of color that you see here, so there tends to be more lighter orange, yellow, and even green, suggesting that the hybridoma antibodies in general hit through different epitope bins; however, the potency of those was quite variable across all of those bins.
0:31:27.2 SK: In this slide, what we’re showing is the clustering trees of paired VH/VL antibody sequences from the B cell cloning method. So in red are the benchmark antibodies, and in green are two of the highest affinity antibodies that came out of B cell cloning. So you can see that most of the antibodies that we pulled out were unique sequences, and when we later compared that with the hybridoma antibodies in terms of sequence diversity, we found that three of the trees had only hybridoma antibodies, and then a separate three trees had only B cell cloning antibodies, and then five of them actually had a combination of both hybridoma and B cell cloning hits. So, all of this taken together suggests that, as far as antibody discovery techniques, it is possible to pull out unique antibodies from each of these different methods.
0:32:27.2 SK: So, this is a summary of all of the characterization data that we generated from the B cell cloning antibodies, and we’re ranking these based on blocking assay IC50 values. And you can see in green are the benchmark antibodies, and so most of the B cell cloning antibodies had equivalent or sometimes even slightly better IC50 values compared to the benchmark antibodies, many of them bound to human PD-L1 by ELISA as well as FACS, and a lot of them bound to Cyno PD-L1 by FACS, although there was variability in the degree of binding with some of these candidates.
0:33:13.5 SK: And again, that might have been because we have not actually screened for Cyno binding on the Beacon Platform, but we could have… That’s something we certainly could have done. We had a majority of these antibodies that did have good binding profiles also had blocking activity, and at least 16 of them had blocking activity that was comparable to the benchmark antibodies, and then two of them had affinity that was in range of the benchmark antibodies at that sub-nanomolar level.
0:33:54.6 SK: In this slide, we’re showing the B cell cloning method compared to hybridoma. Total screened obviously tips in favor of Beacon being able to screen tens of thousands. The time to sequence, now this includes the animal immunization campaign, and so it’s quicker with the B cell cloning on the Beacon Platform rather than waiting for the later stages of the screening process to get that sequence information. The selection of the blockers, on the Beacon Platform, it’s image-based, whereas we used quantitative ELISA.
0:34:37.1 SK: To evaluate blocking activity, we cast a broader net by hybridoma to get 31-100% blocking activity being called out as a hit by hybridoma. Whereas on the Beacon Platform, if it was negative for PD1-Alexa Fluor 488 signal, those were the ones that were selected, so it might have been a more stringent criteria that was used for selecting blocking antibodies.
0:35:08.5 SK: The initial number of blockers was about 35 for Beacon, it was over 40 for hybridoma. We ended up finding good candidates that had high affinity by both B cell cloning as well as hybridoma, but the correlation between having a picomolar affinity and potent blocking activity was tipped in favor of B cell cloning. We had more candidates that had blocking IC50 values either similar to or greater than the benchmark antibodies, and we had greater number of antibodies that had activity, functional activity, in an orthogonal functional assay.
0:36:00.4 SK: So T cell activation assay, we had nine out of 24 of the candidates that had good activity in this assay from B cell cloning, compared to four out of 41 in the hybridoma campaign. And this is an important point because we only screened for one functional outcome on the Beacon, and that was blocking activity, but you’d really like to see that there’s functional activity across a wide variety of functional assays, so this was promising for us to see activity in other functional assays.
0:36:42.1 SK: So in this slide here, what we’re showing is the challenges that… It’s basically a function of hybridoma screening. So again, in the primary screening stage, we cast a broad net, so anything between 30% and 100% inhibition in the blocking assay at the hybridoma supernatant stage was moved forward. But one with a 45% initial blocking, you can see mAb037 there, turned out to be a really, really strong blocker once we evaluated purified antibodies.
0:37:20.0 SK: But there’s other cases where there’s activity that looked really, really good at the primary screening stage, and then at the final screen, the activity appeared to be lost, and that sometimes happens with hybridoma. But luckily a majority of the candidates, so 68% of them, had blocking activity in the primary screen and also showed blocking activity in the final screen as well. But many with the low percentage of blocking activity remained low, so that means that it’s a low-efficiency screening method, and it’s possible that we’re pulling forward candidates that actually don’t have the desired functional properties that we are seeking.
0:38:09.0 SK: So in summary, what we found from our hybridoma versus B cell cloning evaluation was that we had more blockers and high affinity antibodies identified from the hybridoma approach representing 13 different epitope bins. That’s what the LSA data showed us. But unfortunately, none of them were more potent than benchmark antibodies in the blocking assay, and the affinity did not always correlate as well with potency, and fewer had functional activity in an orthogonal T cell activation assay.
0:38:48.6 SK: In contrast, the B cell cloning on the Beacon enabled accelerated discovery of the antibodies, greater sampling of the immune repertoire to identify the potent blocking antibodies. We had, out of those 35, two of them with sub-nanomolar affinity that correlated well with the potency. And we had a greater number of antibodies that had functional activity in the T cell activation assay. And so we think that by focusing in on the antibodies that had that very strong inhibitory activity by that image-based method really enabled us to pull through very strong binders initially on the Beacon platform, and we just don’t have that ability to be able to select candidates from hybridoma with that much confidence because we’re relying on ELISA-based method that can lead to pulling forward candidates that maybe lose functional activity during the subsequent steps of sub-cloning.
0:40:03.1 SK: And so we were very fortunate to be able to take these antibodies from hybridoma and B cell cloning and evaluate them on the LSA. So the LSA’s low sample requirements and speed really aligned well with what we were trying to do here and compare the different antibody platforms. We measured affinity on the LSA that correlated well with the lower throughput techniques, including the Biacore 8K. The network plots were invaluable. We typically do epitope binning by competition ELISA, and we had done that evaluation, in fact. And the relationships between the antibodies, we just did not have the level of resolution that we were able to then see after looking at the LSA data. And just having those network plots really revealed relationships between antibodies with a much higher level of resolution and even hinted at sequence similarities for the hybridoma antibodies.
0:41:09.9 SK: Again, we’re not picking candidates based on sequence yet for the hybridoma antibodies. We basically have to make calls earlier in the process before we even know how diverse the antibodies are, and it’s possible to even pull forward multiple antibodies that have the exact same sequence. And so having a technology available, such as the LSA, can help you to select candidates and get a hint at how diverse they are in terms of the epitope coverage. So it can certainly add value during the supernatant screening stage to select the top candidates
0:41:52.3 SK: And so what we see for future potential is that combining Beacon analysis with the LSA can be really disruptive, so it enables direct and rapid interrogation of the immune repertoire to find rare candidates for challenging targets. And although PD-L1 is not considered a challenging target, it’s certainly not a high-frequency event to pull out an antibody that is sub-nanomolar affinity and has potency in a blocking assay as well as orthogonal functional assays. So we think that we’ll be able to pull out candidates that are really agnostic to therapeutic mechanism of action, therapeutic area, as well as target class, and really select target specific potent antibodies that have the desired affinity, specificity, as well as functional properties earlier in the drug discovery process.
0:42:57.8 SK: So with that, I’d like to thank Carterra again for inviting us to present this story today and for contributing to the data presented here. Also acknowledging Berkeley Lights for their help on the Beacon platform for the proof of concept study. Acknowledge the Shanghai team at ChemPartner for generating all of the reagents and antibodies, as well as characterizing the antibodies by both discovery methods, and also showing the ChemPartner team in San Francisco who’s focused on the B cell cloning effort.
0:43:42.7 SK: Thanks so much for tuning in to this webinar, and now we’re happy to take your questions. We have myself from ChemPartner, Anupam Singhal from Berkeley Lights, and Noah Ditto from Carterra.
0:44:18.0 John: Thank you, Shireen, for that very informative presentation. Now, we will take questions from the audience. If you have any questions, please submit them in the question box on the right. Shireen, was there correlation between epitope bins, as predicted by the LSA network plot and antibody sequence diversity?
0:44:47.3 SK: That’s a great question. So what we found is that, interestingly, there was a correlation between the epitope binning, the antibodies that were clustered together in the same bin and the sequence diversity, suggesting that… And it’s not that shocking because when you actually do those analyses, you kind of predict that similar antibodies would cluster together, but having that level of resolution that the LSA provided did give us a hint. And with the B cell cloning, we already have the sequences, so we could kinda see that pattern emerge. With the hybridoma, we did not actually have the sequence information for that broad set of 41 antibodies, so it was very useful actually to have that clustering and network plot information upfront, because we could potentially pick the candidates that had more potential sequence diversity. So it was a very interesting correlation indeed.
0:46:02.0 John: Thank you, Shireen. Noah, are there technical differences in screening affinity for hybridoma versus B-cell derived antibodies?
0:46:13.3 Noah Ditto: Yeah, good question, John. Generally speaking, no, there would not be… It comes down to considerations about the titers that are in the samples themselves, in other words, absolute antibody concentration, but the matrices they’re in, it really is not influenced in the LSA. Typically in the LSA we capture the clones onto the surface, using maybe an anti-Fc surface, for example. That would effectively do an on-chip enrichment for the experiment, and it means that really any background matrices are washed away before we initiate the antigen titration across that surface to assess binding kinetics. So in short, really any true matrices that came out to the instrument really would not impact the downstream binding kinetics measured.
0:47:11.3 John: Okay, Shireen, do you have access to transgenic rodents to generate fully human antibodies?
0:47:25.1 SK: We have actually done a proof of concept study with the Trianni mouse, very similar to the study done here with anti-PDL-1 as a test case, and we were able to pull out very potent antibodies through that campaign as well. It’s certainly something that works quite well on that platform, and typically for our relationship with any of these companies that have transgenic animals, you typically would have to go get a license for that, and then can come to us and we can certainly execute on that.
0:48:10.2 John: Thank you. Anupam, do you have any other examples of where the ligand receptor blocking assay on the Beacon has been used to down-select lead candidates?
0:48:22.7 Anupam Singhal: Right, that’s a great question. There have been other examples, and the most recent of which is in the current COVID-19 situation. We had collaborators at Vanderbilt University use the assay to identify antibodies that blocked the interaction between the SARS-CoV-2 virus spike protein, as well as the human H2 receptor and that’s a key function in terms of infection of the virus.
0:48:55.2 SK: Okay. Shireen, what are the most challenging aspects of functional assay development on the Beacon platform?
0:49:09.0 SK: The most challenging part is actually how good the reagents are. Our approach is to generate the reagents and QC them by FACS to ensure that we have a robust signal upfront, before even going on to the Beacon platform. If you already have a very good set of reagents and a good read-out by either FACS or ELISA, in general, we’ve been able to transfer those assays onto the Beacon and be able to maintain that dynamic range. Again, it really starts with the reagents themselves, ensuring that you’re actually gonna get that signal off the platform using FACS or some other method, and then, in general, it tends to translate fairly well onto the Beacon.
0:50:04.8 John: Thank you, Shireen. A question about, are you able to enrich B cell population to get a higher hit rate?
0:50:16.0 SK: Yeah, so I can start to answer that and Anupam can maybe chime in as well. Typically what we do is magnetic B cell isolation. We enrich for CD138-positive B cells first, before going on to the platform. It is possible to look for antigen-specific binding upfront as well as another enrichment criteria, but we tend to, in general, like to pull forward all of the CD138-positive cells for mouse, of course; for human, we’re using a slightly different magnetic isolation and enrichment procedure, starting from the memory B cells, but certainly very important to enrich upfront before loading onto the Beacon.
0:51:15.1 AS: Sir. I think Shireen has covered what I would say, and I would just re-echo that the system itself is capable of screening any B cell population, but the better you are able to create a purer secreter population the greater the hit rates that you’re….. using a combination of markers, positive and negative selection on FACS, or whether it’s using standard magnetic enrichment. The system can operate with any of those samples, and obviously performance will be dependent on how effectively the enrichment is performed.
0:52:00.9 John: Noah, in the presentation, there were examples of 8K and LSA data aligning in terms of Affinity, what about on and off rates?
0:52:13.4 ND: Yeah, so the plot did reference affinity, and I think this question is leading towards a concern that when you measure on and off rates, you can have different on and on rates leading to the same measured affinity in the end by varying the two components used as part of the affinity equation. And that’s a good point. There’s actually been a nice manuscript that was put out recently by Arthur Brown at… Adam Arthur was the lead author on it, comparing a variety of things, one of which was kind of on and on rates measured on the LSA and versus the number of technologies, and actually in that comparison it included the Biacore 8K as well, which is referenced here, that’s a nice… There was a really nice correlation of their both on and off rates as well as calculated affinities shown in that data set for a reasonably sized protein antigen. So the short answer would be yes, the expectation is there that you would get comparable off rates between the two platforms once you optimize your assay conditions.
0:53:22.7 John: Thank you. Shireen, do you have a sense of the frequency for false negative blocking antibodies on the Beacon?
0:53:33.6 SK: That’s a really important question. So we actually decided to go back and take 20 of the binding antibodies and moved through the entire process of sequencing, express, purify and characterize, and of those 20, only two of them had binding activity… Well, they all had binding activity, but only two of them had blocking activity, suggesting that we were not losing real blockers early in the process, and even when we compared the blocking activity of those two that we would consider false negative, they were not as potent as the ones that were selected. So it depends on the criteria that you use to select your blocking antibodies, but we set it in a pretty high bar by that image-based method, and if it had a weaker signal there, we just passed on that and then that actually turned out to be less potent antibodies, once we looked at the purified reagents, so we feel that the false-positive rate, at least for this study was fairly low and within an acceptable range.
0:55:00.1 John: Thank you Shireen. We have a question about discussing the improved T cell activation a little bit more, single B-cell versus hybridoma derived antibodies.
0:55:13.5 SK: Yeah, so I hadn’t shown that data, but it was really a function of how many of them had that activity, and there were more that had the functional activity from the B-cell cloning candidates compared to hybridoma. When I refer to T-cell activation, the number that came from hybridoma did have reasonable activity, it’s not that there was a lot of difference in terms of the amount of activation, it was that there were a greater number that came from the B-cell cloning effort.
0:55:56.9 John: Well, thank you. Anupam, are there new features for antibody discovery workflow on the Beacon?
0:56:09.7 AS: Yeah, so since the time that we performed this proof of concept with Shireen and ChemPartner, we’ve launched what we call Opto Plasma B Discovery 2.0, and that workflow now has increased the screening throughput as well as introduced a new on-chip genomics capability where you can perform on-chip seedination to simplify the sequence recovery from B cells.
0:56:44.7 John: Thank you, Anupam. Noah, can you perform epitope binning and kinetics using the same chip?
0:56:58.4 ND: Yes, you can… There’s a little bit of a nuance in kind of how you’re setting it up, particularly because in SDR, you wanna have a sort of optimal density of the antibody on the surface to or the assay you’re attempting, but the short answer is, yes, there’s a number of customers that do like this approach because they can take one censor chip and get to assay two data sets out of the same surface, so they would’ve typically immobilized the antibodies across the surface in a ray format just like Shireen had described in her presentation, and then come in in Kinetics assay, typically up front to assess binding interactions, and then on that same surface come in with a competitive epitope format as well. If there are some cases where maybe the attachment of the antigen might need to be different between the two assays, so that might dictate running a separate experiment but then in cases where the same attachment mechanism is suitable for both approaches, it definitely can be done.
0:58:00.8 John: Thank you, Noah. Carterra has partnered with the Gates Foundation and the La Jolla Institute of Immunology to analyze and screen and characterize therapeutic antibodies. Noah, could you give us a little detail about our work with analyzing therapeutic antibodies for COVID?
0:58:24.5 ND: Yeah, yeah, so for COVID, a lot of the assays kind of described here by Shireen here on the LSA are playing out in the real world in the fight against COVID and trying to find treatments, therapeutic or vaccine-related. So there’s a number of customers that have publicly announced the activities that are using the LSA, this includes AbCellera and Eli Lilly, Distributed Bio as well, to name a few, and they’re all doing very similar assays like this, finding antibodies typically actually from patient-derived sources, doing really detailed characterization, and really the speed of the platform is the key here, because it’s quite obvious to everybody that there’s very limited amount of time here to find a treatment and the world’s population is depending on a treatment in some form or another. So customers have been gravitating to this platform just for the sake of the speed, the throughput, they don’t need much sample to get going, and in the case of Eli Lilly, they already have a candidate in a phase one antibody candidate, which is incredible how fast they turned around from generating a patient derived antibody to actually getting into phase one. And the LSA being part of that workflow, I think is really critical and speaks to its ability to support rapid therapeutic discovery activities.
0:59:54.3 John: Thank you, Noah. Shireen could you please give details about your immunization protocols? What adjuvants do you use? Is there any protocol to increase the number of B-cell clones with higher affinity and recognize native and diverse epitopes?
1:00:12.6 SK: Yeah, that’s a really important question. I can certainly follow up with the details about the immunization strategy. In terms of finding candidates that actually have the higher affinity and native and diverse epitopes, I think it’s all in the screening. So you really have to have the right reagents available to look at whether you’re getting diverse binding to different sites on the protein, for example, but I think really it has to do, you can’t actually rank on affinity on the Beacon, from my experience, Anupam might be able to shed more light on that at this stage, but we are basically looking at bead binding to cells or to beads coded with a particular protein. So we do like looking at bead binding versus cell based binding, ’cause it’s similar to looking at FACS versus ELISA. And of course, it’s just a different presentation of the protein, whether it’s recombinant or expressed on a cell, but having those two things together can be very helpful, and then just being able to screen deeper and evaluate as many as possible, I think is another angle we’ve taken on that, but maybe Anupam can talk more about potentially affinity ranking.
1:01:55.8 AS: Yeah, thanks Shireen. So just on that comment regarding the different assays that you can perform, so we do believe that the combination of the recombinant, the soluble protein assay or on the beads and the cell-based assay will help you discriminate those antibodies that are binding to the native antigen, which is often not expressible in recombinant form. In terms of affinities, we have different customers that have tackled this in different ways, running multiple assays at different antigen concentrations is one way that you can infer how strong a signal it is, but basically… I think right now, the main benefit is that you’re able to select which ones are binding to the native antigen by performing multiple assays.
1:03:01.0 John: Well, thank you Shireen so much for that very informative presentation. We’ve reached the end of the hour, and thank you Anupam and Noah, on our panel for answering these very informative questions. I would like to thank the audience for attending our webinar today. And I hope you all have a really great day. Thank you for attending.