How Can Drug Discovery Programs Benefit From Genetically Engineered Models?
How can success stories become the standard for drug discovery programs?
Success stories have been published and publicized
- but costly failures exist!
Success stories clearly demonstrate that the primary success factor is the physiological relevancy of the model.
Well-designed genetically engineered models can only be obtained through a labor-intensive, science-based analysis process, which relies on three main pillars:
Biology of the target:
Each target is different and belongs to different pathways; engineering should be done accordingly.
Goal of the project:
The model use dictates the design - a model for target validation is different than a model for toxicology studies, even though ultimately the target will be inactivated.
Business constraints of the drug discovery program:
The genetically engineered model is a part of the puzzle and it must fit.
genOway approach: it's all about the right design
For each new model, (1) a thorough scientific analysis is performed (literature, in-silico tools, risk assessment, etc.), and (2) options are ranked and compared, based on the target gene biology, model use, technical feasibility, timelines, and budget.
This well-defined decision-tree process is crucial to the quality of the research model, the quality of data resulting from the use of the model, and its impact on drug discovery programs.
Examples of applications for:
Target validation through tissue-specific expression of the target gene:
Models for ADC: cell-specific expression of the antigen targeted by the antibody, expression level, and distribution are crucial to the design of a relevant model, enabling one to study the effect of ADC on the tumor microenvironment (ADC targeting TAM)
Target validation through gene ablation, the most commonly used strategy for target validation:
Enables one to predict the consequence of a target-complete inactivation, but should be limited to target, with very little interaction with other pathways (ion channels, secreted proteins, etc.)
Target validation through point mutation:
Becoming the standard for dissecting between functions of proteins displaying dual activity, such as scaffolding and kinase activities (RIP1, RIP3, JAK, STAT, etc.)
Mechanism of Action (MOA)
Tissue/time-specific inactivation of a target gene:
Ideal for MOA studies and investigation of targeted therapies (immune check-point, complexes involved in effector functions, and more)
Tissue/time-specific induction of catalytically inactive enzymes/kinases:
To inactivate a given function of an enzyme while preserving its expression and interaction with partners; mimics what a small molecule (DNA sensor, PKC, PK…) would do
The design of the humanized model must take into consideration the family of the target gene and type of therapeutics developed.
Beside its primary function, does the target gene interact with partner molecules? If yes, would a humanization of the target abolish these interactions?
IL-4Rα interacts with the γc chain to create functional IL-4 receptors, but also interacts with IL-13Rα1 to create the IL-13 receptor
Recognition of TLR2 by its ligands requires an heterodimerization with TLR1 and TLR6
Is it required to humanize the whole target gene or shall we humanize only the epitope recognized by the antibody and preserve the interaction with other subunits of a complex?
CD3ε partners with CD3γ, ζ and δ to create a functional CD3 complex
Lack of interaction with partners would results in immune-compromised models
Shall a humanized model express only the canonical form of the target (appropriate for efficacy studies) or the canonical and secreted forms that could act as decoys? (More relevant for PK and PD studies)
The type of therapeutics developed influences the model design:
Use a chimeric molecule to enhance the signaling of the human protein into mouse cells if often suggested but not adapted to GPCR targeted by small molecules since it could modify the binding pocket
Pharmacokinetics and ADMET
What type of models should be used for ADMET studies?
Selection of the species to be used for the study and its genetic background are essential parameters to be taken into account
Easy readouts, early response detection, and enhanced throughput could be obtained using in vivo monitoring of selected biomarkers
How to reproduce the human pharmacokinetics?
Humanization of pathways known to be crucial for PK and PD assessment of biologics and small molecules increases the predictability toward human data: humanization of HSA and FcRN (compound recycling binds HSA complex) closely mimic human pharmacokinetics (hFcRn/HSA off-the-shelf model)
How to detect off-target activities in genetically engineered models?
Physiological relevancy of the model is indispensable to avoid false positives; i.e., most standard KO models exhibit genetic compensation and pathway deregulation
Most studies are done without the appropriate control models, yielding false results and misleading conclusions
Understanding of the etiology of the disease in humans is paramount to designing a relevant translational model.
Are the SNPs identified through GWAS causal agent of the diseases?
What are the consequences of those mutations at the gene and protein levels:
Defect in gene-splicing process resulting in no expression of the protein
Defect in gene-splicing process resulting in expression of a mutated protein
Defect in the gene expression level and pattern
Defect both gene expression level and expression of a functional protein
Answers to these questions are often unknown, but a thorough bio-informatics analysis coupled to in vitro assessment of some of the hypothesis could clarify the context. This dictates the design of the in vitro or in vivo model recapitulating the human situation.
In vivo studies: cost & duration
How can the time to first P.O.C be shortened?
Use of new embryology technologies combined with in vitro phenotyping
State of the art time requested to get first POC data is less than one year (which includes model design and creation, animal production, and POC experiment)
Could the cost of in vivo studies be reduced?
Over the several years of use, well-designed genetically engineered models showed reduced expenses in both human resources and expenses, i.e., less animals per experiments, while meeting ethical considerations
Delivery of cohorts of ready to use, fully validated animals combined with optimized screening methodologies
In vivo monitoring of biomarkers reduces budget in kits and consumables and strongly decreases per experiment cost
On average, and for one target,
at least 10 different types of KO models can be designed (and tens of variations for each type).
Numbers are even higher for more sophisticated models, like Knockin, Humanization, etc.