Fill form to unlock content
Error - something went wrong!
Access the On-Demand Video
Delphine Valente, Ph.D. - Sanofi
Talk Title: Validation of huFcRn Transgenic mouse model to screen novel Fc-engineered monoclonal and multi specific antibodies
Thank you to the organizers for giving me the opportunity to present the validation of humanized transgenic mouse model to screen novel FC engineered monoclonal, but also multi specific antibodies. As mentioned by Adriano the work, I will present you was published in mabs journal. At the end of last year, I will start with the objective of the validation, giving a specific focus on what we know about what drives the clearance of antibodies. And I will follow my presentation with the methodology and the results to finally conclude on the interest of humanized, FcRn mouse model in research and development of biology.
Our main objective was the beginning to build a strategy to support optimization of antibodies in research, driving the selection of the best variance for ultimately predict linear PK in human and also projecting First in human dose, as you may know, is the current gold standard for preclinical PK assessment of antibodies is a non-human primate. Those studies in non-human primate provide a reasonable and early assessment of human pharmacokinetics. However, this species and NHB studies are really costly, challenging to implement in discovery mainly due to the quantity of material needed at this stage. And also because of a lot of ethical considerations. Furthermore, the presence of anti-drug antibodies often precludes the determination of the pharmacokinetic parameters and avoid also the reuse of animals, which is again an ethical problem. This prompted us to work on alternatives there's on in vitro data, such as the use of binding affinity on FcRn and also on, in vivo PK, in transgenic humanized FcRn mouse models.
So among the clearance pathways, just a brief reminder of a lot of for, for you amongst the clearance pathways for antibodies, one of the predominant one is the FcRn mediated, as you probably know, to the neonatal FC receptor as an important function in preventing IgG degradation in vascular on Tutelian sense, receptor-mediated internalization and recycling IgG. However, it could not explain to PKB if your, of all antibodies indicating that med disposition in vivo is really a complex interplay of additional processes besides the FcRn interaction. In addition to non-specific clearance through physico resist and potent resists, also clearance pathways, opioid, such as target mediated specific clearance, but also antidrug antibody mediated clearance. In addition, many of the factors such as antibody properties, like I draw for obesity, but also charges glycosylation pattern, then also impact antibodies clearance. This complexity needed us to consider in first intention, the use of transgenic mice models to recapitulate the different mechanism in order to predict the pharmacokinetic.
So how do we, did we proceed? We first selected a panel of antibodies to test them, and those were same conditions in order to avoid any bias in the analysis we used four diverse backbones as wild type or engineered internally in Sanofi by a group bio research, two kinds of mutations, one twins, humanized FcRn binding to increase elimination half life. You have it in red and in blue into slide. And another one to silence the binding to FC gamma R, which is called NNAS mutation, and which is also a proprietary Sanofi mutation.
We run PK studies on two strengths of transgenic humanized FcRn mass models, the TG 276, which is humanized and also control of the ubiquitous CAG promoter and TG 32 mouse, which is humanized and also control of natural human promoter was being almost equals in parallel. We went pharmacokinetic studies in cynomolgus, male NHP, all animal mouse and monkeys were administered by having those route at the dose sufficiently I to avoid target mediated drug disposition, animals were sampled up to 28 days post dosing and the antibodies were assayed in plasma using a generic LCMS MS Developed, we did also an invitro assessment with measurement of FcRn binding affinity, as well as FcRn FC gamma 3 binding affinity to expand the diversity of the construct tested. We decided also to take the opportunity of compounds, which were in research within Sanofi to add in our analysis to bispecific one wild type and one mutated and one monoclonal antibody with another type of mutation.
So finally, in summary we have at the beginning, a highly diversified data sets for the validation of the model with 16 constructs and immeasurability of multi specific or mutated on FC part antibodies. First of all, year also binding affinity results to FcRn and FC gamma 3 I will not go through all the details, but what I can conclude regarding this data is that regardless the target antigen or antibody format mutations to enhance FcRn binding So mutation ear to endurance FC era, and binding to increase elimination half life showed 1.3 to six fold increase in binding affinity compared to the wild type, in the case of NNAS mutation, the carbohydrate switch mutations completely eliminated the binding to FC gamma as expected, but maintained wild type like FcRn binding protein.
here you have presented the summary of in-vivo inter-species pharmacokinetic parameters, just to show you the ones we focused on, namely exposure, clearance, volume of distribution, and elimination are flat. So we first started to compare the two transgenic mouse model into detach. Your results were controversial on the predictability of both models. And that is, is being often carried out on data, not generated in the same laboratories and different conditions, which can generate the bias. In our experiment. We have clearly observed for TG 276, which are represented in blue dark Intergraph or sharp drop in concentrations for several antibodies tested due to the presence of antidrug antibody. And when no ADA antidrug antibody was present as showed hear force, try AB6 wild-type at the bottom left of the slide, the PK profile was not parallel to Tg32. If you look at the TG 32 at the opposite, the Tg32 in red was parallel to the PK profile in NHB, for all antibodies tested in our panel.
higher clearance and shorter elimination half life was observed in TG 276 compared to Tg32 and NHP then what brought us to the conclusion that the TG 276 is not the best model to predict PK parameters in bigger spaces like NHP and by extension human. Through the entire study, we have observed, like I mentioned before, I incidence of antidrug antibodies in TG 276, but also in NHP, with 33% of NHP, we, I think eight years after single dose, conversely, no antidrug antibodies were observed in our experiment in Tg32 mice, except for one construct, which is MAB as can see the graph with a sharp drop in concentration observed some animals from three days after administration. Nevertheless, as probably, you know, other limit MAB is a monoclonal antibody marketed under the name Humira well-known to be immunogenic and to induce antidrug antibodies in irate of treated patient, despite of its fully human sequence.
Since then we conducted within Sanofi other Tg32 studies with other type of constructs. And we have confirmed the very low incidence of antidrug antibodies with this mice, which gives them a very interesting property to assess pharmacokinetics. In this slide, you have the results of the influence of the FC mutation proprietary to Sanofi, to eminence FcRn binding these results, which were published also in mabs in 2020 show, clearly that this novel mutations increase elimination half life of bevacizumab, which is a map tested here in comparison to the wild type. The wild type is in black and the mutated one corrode. And we see that extension of the half life is in the same extent as the well-known LS mutation in both trend on the left path, TG 32 mice PK profile and on the right, you have the NHP sorry. I have something right here, which is not normal.
But anyway, here you have the first results on the influence of the FC mutation to enhance this FcRn binding no longer on a mono specific antibody, but on the tri specific antibody. And as you can see on the graph there, the elimination half life of the mutated constrict, which are red, is clearly increased in both spaces. Again, you have the Tg32 mice on the left and NHP on the right, which leads us to conclude that the Tg32 mouse model is an appropriate one to distinct reach pharmacokinetics differences, not only in mono specific antibody, but also in multi specific antibodies.
And also interesting observation we made during this analysis was on the specific case of a monoclonal antibody. As you can see on the graph in both species on the left side tg32 on the right side, NHP. The NS Mutation in red does not increase in this case elimination half life, of the antibody, despite a five fold higher in vitro binding affinity of these LS mutated antibodies to human eyes. FcRn nevertheless the wild-type antibody have already a long elimination of life compared to all the monoclonal antibodies. And in this is case, it seems that in this it's difficult to further increase the elimination half life with the FC mutation.
We have seen the influence on pharmacokinetics of different mutations to enhance H U FcRn binding and increase the elimination half life. Now, in this slide, we have the results of the impact of NNAS mutation, which is a mutation which silence the FC gamma binding on the PK profiles of antibodies. We did a comparison for two monoclonal antibodies first for mAbs, which is in green, here in the, in the graph, the wild type, which is in, in the not dotted line. So mutated is in dotted line and the bevacizumab right type and mutated in red in Tg32 mice and in NHP. And again, we clearly observed here a super imposition of wild-type and NNAS pharmacokinetic profiles for the two antibodies in both species, consequently TG32 mouse is well treated, two screens, absence of impact of the NNAS mutation on the PK profile of monoclonal antibodies as seen in non-human primates.
Here You have to compare reason of linear clearance of monoclonal on multi specific antibodies with or without FC mutation. And we see interestingly that multi specific's exhibit a much higher clearance compared to monoclonal antibodies, two, two, five, four, two, two, five fold. And this trend observed in nonhuman primate was exactly the same in Tg 32 mouse showing again, the great interest of this transgenic mouse model to evidence changes in clearance of multi specific antibodies. We tried to establish a relationship between linear clearance in non-human primate and Tg32 mice of diverse antibodies tested and in vitro binding affinity to human FcRn. And as you can see on the graph, we add a poor correlation, even if this in-vitro assay as demonstrated utility, for sure, for identifying PK risk liabilities, a strong correlation cannot be established as it is not the only factor which impact the linear catabolic clearance commentarial approach are under evaluations within our company, as it is likely to confer the greatest predictive power, but to date, we are not.
In this slide is presented the coalition of TG 32 and NHP half life and clearance half life on the left and side and clearance on the right side of the slide, we can observe a significant correlation of the enemy nation half life between Tg32 models. And non-human primate for the panel of antibodies tested. Overall, we have observed also a good correlation for Clarence, but as you will see later on, we need a scaling factor to scale from Tg32 to NHP clearance. So in conclusion, we can say again, that there is an excellent potential for the TG32 mice to be used instead of non-human primates for lead optimization, not only for wild type monoclonal antibodies, as it has been demonstrated in the literature, but also for FC engineered monoclonal antibodies bispecific and trispecific. The practice of projecting monoclonal pharmacokinetics in larger species and in human is well established and general approach for antibody with linear PK, commonly used is empirical allometric scaling that is based on the power relationship between body weight and PK parameters, volume of distribution or clearance.
The scaling exponent has been established and published for monoclonal antibodies, but again, nothing was available in the detail at IU for multi specific and FC engineered. Our work here has been to use our data from TG 32 mice, and NHP to establish this center metric exponent two scales is PK parameters for wild type and compare with Muti specific. And we have observed that exponent are not so different for clearance exponent as been determined to be 0.91 For Fc mutated antibodies and multi specific in comparison with 0.97 For wild type and for volume of distribution, it is 0.93 in comparison with the whole 495.
We used this exponent determined for clearance and volume distribution to predict the pharmacokinetic parameters of our data set of constructs. And as you can see here, we have a good prediction with all predicted values within two fold of the observed values. This gives us a method to predict from Tg32 mouse, the PK parameters of the non-human primate and by extension of humans. And now I would like to summarize and to conclude on the interest of the TG32 humanized FcRn mouse model in research. Further it's a confident and cost effective model to facilitate the screening strategy, to select the antibodies. As you have seen, it's a highly sensitive model for differentiating PK of variants of FC engineered, monoclonal antibodies, but also multi specific antibodies. This model is well-suited to screen the absence of impact of silencing FC gammaR mutation on the PK profiles of antibodies further more.
This is a good model to accurately predict cleaner part of the pharmacokinetics in NHP's using the scanning factors established in your publication and by extension in human. This will be very, very useful for guiding the design of further pharmacology and safety studies in research and development, the low immunogenicity of humanized antibodies in the Tg32 mouse model with a very low incidence of ADA incorporation with NHP allows a prediction of PK profile in human, in the absence of a reliable NHP data due to antidrug antibodies. Furthermore we can enable also the determination of potential target mediated or disposition at lower dose in NHP. If you have observed data showing much higher clearance than the predicted one, you can imagine that you have a TMDD in your NHP and last but not least can provide you information about linear and non-linear components of the clearance in NHP enable further development of mechanistic PKT MDD models for more reliable PK extrapolation. And I have finished my talk. I would like to thank you all for your attention and to all the main contributors of this work, and I'm ready for questions.