Bias Translation: Possible Reasons for Failure
In terms of translation of biased agonism found in human cells in vitro , there are two considerations. The first relates to drug development in that a given pharmacologic profile must be linked to a potential therapeutic indication and often that is done through testing in animal models. Therefore, this involves the transfer of bias from a human receptor to an animal orthologue receptor. For orthosteric ligands this usually is not a serious impediment as residues required for natural ligand binding and structural integrity of receptors are highly conserved whereas residues critical for allosteric signaling are poorly conserved ( Leandera et al, 2020). In general, orthosteric natural agonist recognition sites for families of receptors with a common agonist (such as the 5 member family of acetylcholine receptors) is difficult due to the similarity of the conserved acetylcholine binding sites (Gentry et al, 2013; Myslivecek, 2022). However, since bias involves protein allostery, there are data to suggest that bias transduction could be more sensitive to receptor structure and a.a. homology than standard orthosteric agonism even through nuances of agonists binding differently at the same binding site (metabotropic glutamate receptor 5 (Hellyer et al, 2020). In fact allosteric effects have been shown to be particularly sensitive to minor differences in a.a. homology possibly because the residues involved in allosteric function go beyond those involved in agonist recognition. Therefore, differences in residues distant from the natural agonist binding site can be critical to the effect of a ligand produced at the receptor binding site and single amino acid mutations have been shown to produce serious effects on allosteric molecular function. For example, the muscarinic receptor positive allosteric modulator BQCA produces >10 fold potentiation of acetylcholine effects in native receptors but a single amino acid mutation (Y3816.51A) in the receptor completely negates the effects of BQCA (Abdul-Ridha et al, 2014). Similar differences in biased signaling are seen with the dopamine D1 receptor agonist bias for agonists at dopamine D2 receptors. While in the Wild Type Dopamine D1 Receptor DPAT D1 is biased toward ERK, a single a.a. mutation reverses DPAT bias toward cAMP ( Tschammer et al, 2011). Differences in a.a. sequence, as seen with human and animal receptor orthologues, can show large differences in allosteric modulators as in the case of allosteric potentiating modulators of the glutamate receptor Type 1; these show large differences in activity between human and rat receptors (Cho et al, 2014). Since bias is allosteric modulation, this may pose special problems testing translation of in vitro human signaling bias to animal therapeutic models to assess significance.
The second issue with bias translation does not inolve species differences but rather, translation across different human systems. Progress in new technologies has brought vast improvements in screening hit rates and the development of new molecules but in spite of these advances, the success rate of actual drug candidates that are useful therapeutically is still surprisingly low. One estimate suggests that 50% of all new drug candidates fail because of lack of efficacy (Arrowsmith, 2011) where here efficacy is defined as the candidate performed as required in human therapeutic settings. Aside from commercial and safety issues, this rate of failure indicates a serious shortcoming in the drug discovery process in that it represents the fact that, after rigorous state of the art application of pharmacology and discovery science, seemingly optimal candidate molecules still do not do what they were supposed to do in humans. Literature analyses suggest that the difficulties may be related to failure to verify compound exposure and to demonstrate physical target engagement in the relevant therapeutic tissue (Morgan et al, 2012; 2018; Bunnage et al, 2013; Cook et al, 2014) but with biased ligands, there may be other issues. Obvious reasons for miscalculation are failure to recognize what efficacy is needed to treat the disease, and/or a wrong choice of biological or chemical target. However, another factor that may be under-estimated is a failure to adequately characterize the true efficacies of the candidate molecule; this may especially be true in the case of biased molecules. From this standpoint the question could be asked, is it enough to characterize receptor-mediated bias for a candidate molecule without further classifying possible texture in biased signaling in the cell?
There are a number of possible dissimulations between initial bias estimates at the receptor level and complex in vivo signaling profiles; it is worth considering these. An in vitroidentification of biased signaling furnishes a premise that the ligand in question will produce differential signaling compared to the natural agonist. Historically, initial ideas on how bias could improve therapy (see Kenakin, 2019) were gained from simplistic comparison of agonism in two quantifiable assay systems, cyclic AMP and β-arrestin. Presently there are several more sophisticated analyses to predict favorable prospective biased signaling than previously considered when only G protein and β-arrestin signaling were the options (vide infra ).