Unknown Downstream Outcomes from Receptor-based Bias
While bias most often is detected at the level of the receptor, the stabilized receptor active state conformation may code for unknown events further on down into the cytosol. In light of the complexity of signaling in cells, it is possible that the benefcial effects of a biased agonist may be lost in the milieu of signals and biochemical cascades. Dependence on a single readout of bias at the receptor level may not identify molecules that produce unique effects further down the signaling cascade. For example, BRET association of receptors with β-arrestin indicates increased association but does not necessarily augur receptor internalization and/or signaling (vide infra ). Recent data with β-arrestin suggest that heterogeneity in β-arrestin conformations may lead to trafficking of stimulus in various ways after receptor-β-arrestin interaction (Chen et al, 2023). Specifically, while BRET analysis of receptor association with beta arrestin may be found, it may be critical which beta arrestin confromation is involved (i.e. ‘tail’ or ‘core’, (Cahill et al, 2017) to further delineate internalization vs internalization for the formation of a ‘supercomplex’ of the receptor and β-arrestin in endosomes providing sustained signaling (Chen et al, 2023). Therefore, a program seeking an internally signaling agonist for prolonged cellular activity might test compounds for preferential β-arrestin bias (assuming β-arrestin is the vehicle for transport to the endoplasmic reticulum) but further experiments might discern agonist activity toward that desired endpoint. For example, fig 3 shows that agonist 11 is identified as the most biased for β-arrestin. However, further testing could be done at this stage to differentiate which β-arrestin conformation is preferred; in this case, it may be that agonist 11 would not be the preferred compound as it is not biased toward the predicted β-arrestin conformation that could lead to ER signaling. In general, testing sub-groups of biased agonists may further characterize useful activity, especially in cases where an initial testing of a biased agonist does not provide a more informative outcome. Follow up studies with other biased molecules could provide a clearer answer; as shown in the figure, identfication of agonist 11 as an exemplar biased agonist in this case would not yield the required signaling as the stimulus bifurcates throughout the cell whereas agonists 13 to 16 might have provided a better choice.
Bias translation involves the synoptic nature of pharmacologic response (i.e. the necessary partnership of the activated receptor with complex signaling patterns) and brings into consideration the cellular milieu of the active state receptor-cell mixture. In addition to the production of a variety of ligand-bound receptor active states, comes the subsequent interaction of these states with another constellation of (for example) possible β-arrestin conformations. Specifically, the conformational flexibility of β-arrestin allows GPCR-induced conformational rearrangement to expose distinct binding surfaces that allow recruitment of different effectors for specialized signaling complexes (Haider et al, 2023; Luttrell et al, 2018; Shukla et al, 2014). This can lead to considerable heterogeneity in cytosolic signaling and factorial combinations of outcomes are possible. Possible other dissimulations with respect to expected therapeutic outcomes at this stage include:
  1. Different signaling partners in different cell lines Agonist activity on receptors transfected into different cell lines have shown differences in relative agonist potency. For example, transfections of calcitonin receptors into two different host cell lines (CHO cells, COS cells) show large differences in the relative potency for porcine Cal, human Cal and h CGRP. Specifically, in CHO cells the relative potency is hCGRP/hCal/pCal of 1/ 10/500 whereas in CHO cells the potency ratios are 1/2/8 (Christmanson et al, 1994). Total synoptic agonist response can be revealed through label free assay formats such as cell impedance; these have been to used to measure the relative potency of dopamine agonists in two types of cells (U-2 cell, SK-N-MC cells) where the cell type made a 4-fold difference in the relative potency of agonists dopamine and A77636 (Peters and Scott, 2009). This translates into differences in cell bias as function of cell type
  2. Variant stoichiometry between receptors and signaling components . The most obvious variable operative here is the relative stoichiometry of receptors and signaling components in various cells. The relative stoichiometry of receptors and signaling proteins is a well established variable in functional pharmacology and this brings into play the role of the host cell in bias measurement and detection. Thus, a paucity of an important signaling partner (i.e. G protein) could negate a bias seen in a system where this is not the case i.e. Eason et al, 1992. This is particularly relevant to low efficacy agonists where response could disappear with low signaling coupler in a given cell; this may be a factor in the bias seen with the biased opioid receptor TRV130 (Singleton et al, 2021). An even more surprising effect can be seen for truly biased agonists in receptor systems without limitations in coupling proteins. For example, the relative potency of the full calcitonin agonists eel and porcine calcitonin for human calcitonin receptors transfected into in wild type HEK 293 cells is EC50(eel Cal)/ EC50 (pCal) = 0.4; Co-transfection of Gαs protein to enrich the natural Gαs content produces a complete reversal of the relative potency of eel and porcine calcitonin. In this enriched cell line, the relative potency is reversed to EC50 (eel Cal)/ EC50 (pCal) = 8 (a 32-fold difference) (Watson et al, 2000). Reduction of key signaling components in a cell clearly can limit low efficacy agonists from utilizing pathways but a recently interesting variation on the theme of receptor-signaling protein relative stoichiometry suggests that actual increases in cellular receptor can affect the observed bias (Li et al, 2023). Differences in receptor expression levels in cells also can introduce a temporal dissociation for response as in the case of GPR84 where a delayed and suppressed activation of Akt was found in low expressing cell lines (Luscombe et al, 2023).
  3. Variant Stoichiometry of GRKs and β-Arrestin: There are striking cases of ligand-directed signaling to the β-arrestin system through targeting GRKs. For example, the chemokine receptor CCR7 has two natural agonists (CCL19, CCL21) and while CCL19 leads to receptor phosphorylation and β-arrestin recruitment through GRK2 and GRK6, CCL21 activates only GRK6. This differential activity on GRKs leads to different cytosolic consequential responses for these two chemokines (Zidar et al, 2009). Other studies indicate with receptor structural studies that the interaction of the neurotensin 1 receptor with GRK2 guides the receptor-β-arrestin interaction for signaling (Duan et al, 2023). GRK may not be the only player as β-arrestin/GRK complexes also may require the presence of ubiquitin to direct receptor selectivity (Liu et al, 2023). In general, data with GRK knockout cells reveal the importance of GRKs as major players in the cytosolic signaling of GPCRs (Drube et al, 2022) therefore variation in cellular GRK levels can modify GPCR signaling from initial in vitro bias determinations in other cell lines.
  4. Varying temporal differentiation of agonist signals in cellsIn vitro assays are snapshots in time with no real regard to the timescale of real life physiology but rather are optimized for accuracy of measurement. The two main in vitro assays used to detect bias (second messenger G protein vs β-arrestin) have very different timescales for steady-state and maximal effect making comparisons possibly dependent on when the measurements are made. For example, a study of dopamine D1 receptor agonist bias on cyclic AMP and β-arrestin shows a discernible temporal difference thus introducing a possible dissimulation in the assessment of signaling bias (Klein Herenbrink et al, 2016). These temporal differences extend to the cell where G protein activation of ERK is rapid and transient and β-arrestin activation of ERK are more sustained involving translocation to the nucleus (Liu et al, 2023). Temporal dissociations extend beyond short acute timespan responses to the production of transcription effects leading to protein expression. In general, the time bias measurements are taken yield somewhat arbitrary indices of differential signaling that may have unpredictable effects in real time physiology. These dissimulations in time emanate from the cellular translation of receptor activation and not necessarily from the timescale of ligand-receptor interaction. In fact it has been shown that the bias of opioid agonists are independent of the rate of interaction of the molecules with the receptor (Pedersen et al, 2020). New techniques are being used to explore this variant in bias, i.e. genetically-encoded fluorescent biosensors have been employed to illuminate spatiotemporal biased signaling (Kayser et al, 2023).
  5. Location bias: Agonists that target receptors to β-arrestin leading to internalization can show bias with respect to the location (and function) of the internalized species (Eiger et al, 2022;Wang et al, 2023). For example, the chemokines RANTES and AOP-RANTES both internalize CCR5 receptor (for prevention of HIV-1 infection) but whereas RANTES internalizes the receptor which then rapidly re-emerges through recycling, AOP-RANTES internalizes the receptor to shunt it to lysosomal destruction (Mack et al, 1998). Thus the actual receptor conformation stabilized by the agonist may determine the fate of β-arrestin bound receptors. Location bias also has been noted for GLP-1 agonists where in studies all GLP-1 agonists activate nuclear ERK1/2 activity but the agonists liraglutide and oxyntomodulin (biased towards pERK1/2 relative to cAMP when compared to GLP-1 and exendin-4), show spatiotemporal control by also stimulating pERK1/2 activity in the cytosol (Fletcher et al, 2018).
  6. Variation of the magnitude of bias significant with respect to the overall cell response While ‘bias’ can be detected in in vitro systems, there is no guide as to the significance of that bias to whole body physiology. Bias indices can range from fairly modest (2-3 fold) to values >10-20 raising the question, what level of bias is physiologically significant? While this probably will be system dependent, a measure of how powerful apparently small bias values can be is demonstrated by diazepam, an anxiolytic with known prominent therapeutic activity. Specifically, diazepam produces a mild two-fold sensitization of GABA response but this translates to an 80% increase in GABA response and a well-known significant physiological effect (Skerrit and MacDonald, 1984). This suggests that bias values of 2 or greater might have a significantly affect on agonist phenotypic activity. Calculation of the bias of the opioid analgesic TRV130 for cyclic AMP over β-arrestin using ΔΔLog(max/EC50) values (Kenakin, 2017) yields a value of 3.39 (Singleton et al, 2021), ostensibly a relatively low value but of possible significance in light of data with diazepam. It should also be noted that the low efficacy of TRV130 for β-arrestin signaling (TRV130 has 33% of the efficacy of morphine for G protein and only 15% of the efficacy of morphine on β-arrestin) may significantly contribute to the beneficial profile of this molecule. Considering a relatively ‘balanced’ agonist that generally does not distinguish signaling proteins versus a biased agonist that does, the question arises does the degree of bias influence variability in terms of translation (i.e. variation in potency with differences in cell type and/or tissue sensitivity)? A theoretical model of a single receptor interacting with two coupling proteins (Kenakin, 2003) (Fig 4A) indicates that the potency ratio of two balanced agonists (or two agonists of identical bias) will not deviate in two cells lines of varying G protein make-up (CellA = [G1]/[G2] = 1 ; CellB = [G1]/[G2] = 10; Fig 4B). In contrast, for two agonists of different bias , the difference in G protein composition can produce radical differences in relative potency (Fig 4C). It can be seen that the relative potencies of non biased agonists remains the same whereas the relative potencies of the agonists of different bias actually reverses with the change in G protein composition. These simulations suggest that the degree of bias may contribute to the variability of translation of agonist effect in different cell types.
  7. Different levels of assessment of bias within the stimulus-response cascade in the cytosol: When considering bias it is also relevant to think about where in the cytosolic signaling cascade the bias may make a difference. Standard in vitro bias assays generally assess differences at the receptor level but as signals bifurcate throughout the cell, the emphasis on discrete pathways may vary. For example, Fig 5 shows the effect of seven dopamine agonists measured from the point of view of six response pathways. When bias is assessed for each pathway through ΔΔLog(max/EC50) values, it is interesting to note that the magnitudes of the bias indices vary with pathway indicating that as the signal propagates from the cell, it is differentially modified in an agonist dependent manner (Klein Herenbrink et al, 2016). These types of effects reveal the texture of bias as a function of the number of vantage points used to make the measurements. For example, G protein selective PTH analogs build bone through cAMP and β-arrestin selective analogs would not be predicted to be as efficacious since the β−arrestin activity terminates G protein signals. However, paradoxically, β-arrestin selective analogs also build bone mass in vivo largely through regulation of cell-cycle, survival and migration/cytoskeletal dynamics (Luttrell et al, 2018). Arrestin-focused response signatures can further be explored through arrestin-dependent transcriptome signatures to predicted outcomes of biased agonism (Maudsley et al, 2016).
Improving Bias Translation:
There are numerous theoretical and practical hurdles to the accurate translation of in vitro bias to complex in vivo systems raising the question, how can these be minimized to optimally design biased agonist programs for success? The first step is to identify bias in a molecule and this can be done through cross-screening in two assays and comparing the results with a bias plot. Considering the complexity of allosteric differences with different receptor conformations, it probably is not too important which two pathways are chosen; G protein signaling and β-arrestin historically have been the standards. The main function of this first step is the identification of a candidate biased molecules which may produce a useful agonist phenotypes in vivoby stabilizing unique receptor conformations. However, out of an array of biased candidates, their ‘robustness’ in terms of resilience of bias to varying cellular conditions could be tested:
  1. Identify ‘Efficacy-based’ over ‘Affinity-based’ bias : Bias based on differences in efficacy (Rajagopal et al, 2011) are more resilient to changes in cell sensitivity than those based on affinity.Experiment : Test bias for immutability in cells of varying sensitivity (i.e. varying levels of receptor expression).
  2. Measure the intrinsic efficacy of the candidates in single pathways : Low efficacy in a negative pathway can be useful to strengthen bias under a range of in vivo conditions and a measure of agonist efficacy can be obtained through manipulation of levels of receptor expression (Jiang et al, 2022). Experiment : Measure the relative efficacy of the candidate to the natural agonist with the operational model.
  3. Measure variation of bias in different cellular backgrounds:Bias could be measured in a range of host cell lines to gauge variation in biased signaling. For example, GRKs have been shown to affect signaling bias and cells have variable GRK levels and variable levels of GRKs can be used as a variable to assess the impact of GRK levels on agonist bias (Matthees et al, 2021). Experiment:Assess bias in cells with varying levels of GRKs and/or varying cell types.
  4. Measure the temporal dependence of bias estimates: Some agonists yield time dependent estimates of bias which could make them unstable predictors of in vivo bias; for instance, while dopamine cAMP/β-arrestin bias is stable when measurements are made over 90 min, aripiprazole changes by a factor of 10 (Klein Herenbrink et al, 2016). Experiment: Measure bias at two separated timepoints.
  5. Apply more textured estimates of signaling heterogeneity:While simple assays such as cyclic AMP and β-arrestin BREThave been used to good measure in this field, the availability of first line assays to further differentiate active state signaling can offer advantages. Thus while cyclic AMP may augur effects of the agonist-activated receptor on Gαs signaling, assays that differentiate all G protein signaling (such as TRUPATH, (Olsen et al, 2020) may offer rapid first-line separation of agonist profiles.Experiment: Utilize more textured G protein assays (Soave et al, 2020) as the first differentiator of bias.
  6. Measure agonist receptor off-rates: Slow dissociation of molecules from receptors can cause pharmacodynamic-pharmacokinetic dissociation and favorable in vivo target coverage. This is a property separate from potency and/or bias but is essential in the characterization of a future in vivo candidate.Experiment: Measure agonist off-rates.