Abolishing β‐arrestin recruitment is necessary for the full metabolic benefits of G protein‐biased glucagon‐like peptide‐1 receptor agonists

Earlier studies have shown that peptide glucagon‐like peptide‐1 receptor (GLP‐1R) agonists with reduced β‐arrestin recruitment show enhanced anti‐hyperglycaemic efficacy through avoidance of GLP‐1R desensitization. However, the ligand modifications needed to decrease β‐arrestin recruitment usually also reduces GLP‐1R affinity, therefore higher doses are needed. Here we aimed to develop new, long‐acting, G protein‐biased GLP‐1R agonists with acute signalling potency comparable with semaglutide, to provide insights into specific experimental and therapeutic scenarios.


| INTRODUCTION
Glucagon-like peptide-1 receptor (GLP-1R) agonists are mainstay treatments for type 2 diabetes (T2D) and obesity, improving blood glucose levels through effects on β-cell function and insulin sensitivity via weight loss.The GLP-1R is a class B G protein-coupled receptor (GPCR) highly expressed in metabolic tissues including pancreatic β-cells and other islet cell types, anorectic neurons in the peripheral and central nervous system, and at lower levels in myriad other tissues including kidney, myocardium, blood vessels and immune cells. 1 All currently approved GLP-1R agonists are synthetic analogues of the endogenous agonist GLP-1(7-36)NH 2 or its paralogue exendin-4, with a variety of modifications that extend pharmacokinetics through proteolytic stabilization and/or avoidance of renal clearance. 24][5][6][7][8] In many cases, this results in diminished coupling to intracellular effectors, with recruitment of β-arrestins affected more so than G protein-dependent cyclic adenosine monophosphate (cAMP) signalling.This imbalance is often classified as 'bias' in favour of G protein-dependent signalling.A consistent effect of G protein-biased GLP-1R agonists is their prolonged GLP-1R signalling ability, circumventing natural mechanisms that prevent receptor overstimulation such as steric hinderance by β-arrestins and receptor internalization. 4,5,9e probable importance of biased agonism has been highlighted by the realization that the GLP-1R/GIPR co-agonist tirzepatide, recently approved for the treatment of T2D, shows pronounced G protein bias at the GLP-1R. 10However, the acute cAMP signalling potency of tirzepatide at the GLP-1R is reduced substantially compared with semaglutide, 11 potentially accounting for the higher doses required clinically (up to 15 mg vs. up to 2.4 mg, respectively).An earlier long-acting, acylated G protein-biased GLP-1R agonist developed by our group also showed reduced acute signalling potency compared with the non-biased comparator, although it still showed better anti-diabetic effects in mice. 9Feasibly, restoring acute cAMP signalling potency may further add to the benefits of G-protein bias.In addition, it is unclear whether total abolition of β-arrestin recruitment (for maximum avoidance of desensitization) is preferable to a partial reduction, as β-arrestin recruitment is paradoxically linked to positive signalling effects including GLP-1R-stimulated insulin release. 12ke all proteins, the GLP1R gene sequence is subject to natural coding variation.4][15][16] The importance of target coding variation on therapeutic and side effect profiles for GPCR-targeting drugs has recently been proposed on theoretical grounds. 17The effects of GLP1R missense variants on signalling characteristics of GLP-1, oxyntomodulin and exendin-4 were previously tested. 18,19However, the impact of coding variation on G protein-biased GLP-1R mono-agonist signalling is not yet known, and, because of larger differences in ligand signalling properties, may be more consequential.Understanding how target variation influences responses to GLP-1R agonists will be necessary for personalized treatment selection in T2D. 20 this work we integrate our previous knowledge of ligand factors responsible for G protein-biased agonism to develop new, pharmacokinetically optimized GLP-1R agonists with reduced coupling efficiency to β-arrestin recruitment but retained cAMP signalling potency.We then use these ligands to address questions pertinent to

| MATERIALS AND METHODS
New GLP-1R agonist peptides were assessed using a variety of in vitro and in vivo assays (full methodological details are provided in Data S1).

| Development of equipotent Gα s -favouring glucagon-like peptide-1 receptor agonists
Our previous development of biased GLP-1R agonists showed that substituting phenylalanine for histidine at the N-terminus of exendin-4 markedly reduced recruitment of beta arrestins, while moderately reducing acute signalling potency for cAMP. 4G protein-biased signalling properties were maintained with a C16 fatty diacid chain installed at the peptide C-terminus for half-life protraction, but overall binding affinity was further reduced. 9,21Therefore, here we introduced further targeted amino acid sequence substitutions to improve peptide signalling potency and stability, in addition to optimizing the acyl chain and linker.After iterative improvements we arrived at two peptides, SRB106 and SRB107, which share a C20 fatty diacid chain at the C-terminal lysine residue and differ only at their N-termini (Figure 1A; Figure S1A).Various linkers and acyl chains were tested, and the εLysγGlu motif was identified as providing an optimal balance of half-life protraction and potency.Like semaglutide, SRB106 and SRB107 were GLP-1R mono-agonists for both human and mouse receptor cAMP responses in transiently transfected HEK293-derived T-REx 293 cells, with no meaningful activation of the glucagon receptor (GCGR), glucose-dependent insulinotropic polypeptide receptor (GIPR) or glucagon-like peptide-2 receptor in either species (Figure S1B, C, Table 1).
To seek evidence of biased agonism, we used the NanoBiT system to evaluate ligand-induced coupling to mini-Gα s (a conformational sensor for G protein-favouring active GPCR states 22 ) and β-arrestin-2 recruitment.Assays were performed in parallel using T-REx 293 cells stably transfected with SNAP f -tagged human GLP-1R with a C-terminal SmBiT tag (referred to hereafter as SNAP-GLP-1R).SRB106 and SRB107 both showed reduced maximum engagement with mini-Gα s and β-arrestin-2 compared with GLP-1(7-36)NH 2 and semaglutide, with SRB107 most affected (Figure 1B).Notably, the β-arrestin-2 recruitment response for SRB107 was barely detectable.To compare formally the relative coupling of each ligand to these responses we used the modified operational model, 23 setting mini-Gα s as the reference pathway.This indicated enhanced cAMP signalling for SRB107 compared with semaglutide and SRB106 (Figure 1C).Of note, pronounced loss of the β-arrestin-2 response with SRB107 was also seen in cells transiently expressing mouse or human untagged GLP-1R (Figure S1D), and both β-arrestin-1 and β-arrestin-2 recruitment were comparably reduced when tested using the PathHunter GLP-1R system (Figure S1E).
To assess the signalling performance of each agonist at natively expressed GLP-1R, we also used the sensitive, fluorescent cAMP biosensor cADDis, 24 transduced into dispersed primary mouse islet cells, rat-derived INS-1832/3 insulinoma cells 25 and the recently developed human β-cell model EndoC-βH5. 26Stepwise addition of increasing concentrations of agonist was used to assess cAMP responses across a wide concentration range by fluorescence imaging (Figure 1D).
These experiments indicated that semaglutide, SRB106 and SRB107 show similar maximal responses and potency at physiological GLP-1R expression levels, except for a modestly reduced potency for SRB107 in mouse islet cells (Figure 1E, Figure S1F, Table 1).We also found that 100 nM of each test agonist led to a similar to increase intracellular calcium concentrations in intact mouse islets at 6 mM glucose (Figure 1F).
Overall, these data indicate that semaglutide, SRB106 and SRB107 are approximately equal potency GLP-1R mono-agonists for cAMP, but with the latter two showing partial agonism for β-arrestin recruitment.The partial versus complete loss of β-arrestin recruitment with SRB106 and SRB107, respectively, along with relatively preserved cAMP signalling potency, provides an opportunity to examine the optimum degree of β-arrestin response attenuation for maximum metabolic benefits.

| Glucagon-like peptide-1 receptor internalization is slower with SRB107
G protein-biased GLP-1R agonists with reduced β-arrestin-2 recruitment typically show reduced GLP-1R internalization.Reduced uptake of 100 nM SRB107-TMR was confirmed quantitatively using an acid wash protocol to remove residual surface-bound ligand (Figure S2E).We noted in the latter study that semaglutide-TMR showed unexpectedly reduced uptake, despite its evident endosomal distribution when imaged at high resolution; this will require further studies to resolve.In addition, after pre-treatment with 100 nM unmodified SRB107, there was more GLP-1R available at the cell surface for subsequent binding of fluorescent exendin-4-Cy5, 6 suggesting less receptor had been internalized (Figure S2F).
To identify any potential species-specific agonist response differ- These experiments all pointed to reduced GLP-1R internalization with SRB107, although we note this effect appeared less marked at the mouse compared with human GLP-1R.Finally, to assess internalization effects with endogenously expressed GLP-1R, we imaged primary mouse pancreatic islets after TMR agonist treatment, which again showed predominant membrane retention of SRB107-TMR (Figure 2E).
These experiments confirm a specific reduction in the ability of SRB107 to promote GLP-1R internalization.
T A B L E 1 cAMP signalling potency estimates. GLP

| Sustained stimulation effects in the presence of physiological albumin concentration
Acylated GLP-1R agonists are extensively bound to albumin, reducing free drug concentration in plasma.Indeed, compared with experiments conducted at 0.1% bovine serum albumin (Figure S2B), the presence of a physiological concentration of human serum albumin (HSA; 4%) led to a right shift in cAMP signalling potency for semaglutide, SRB106 and SRB107 in the same cell type (Figure 3A), with EC 50 estimates of 0.4, 0.5 and 0.6 nM (Figure S2B) shifting to 16.2, 6.9 and 14.0 nM (Figure 2A), respectively.We note that this right shift is less marked than the HSA-induced shift in binding affinity noted during the development of semaglutide, 28 which is not surprising as the cAMP pathway is highly amplified.
To probe whether GLP-1R internalization effects would probably be operative under conditions mimicking in vivo drug exposure, we measured the effect of sustained agonist stimulation on the net loss of surface SNAP-GLP-1R in T-REx 293 cells in media supplemented with 4% HSA.Here, at the highest concentration tested (1000 nM), 12 h stimulation with semaglutide or SRB106, but not SRB107, led to marked SNAP-GLP-1R downregulation (Figure 3B).Importantly, differences in the extent of surface GLP-1R loss with each agonist were observed the reported range of steady state plasma semaglutide concentrations achieved with 2.4 mg weekly dosing in humans, 29 and for mice after a single 10 nmol/kg dose as in our study. 30We also assessed the same phenomenon in INS-1832/3 cells and intact mouse pancreatic islets using post-labelling with exendin-4-Cy5 6 ; here, with 4% supplemental HSA treatment, 100 nM SRB107 again led to moderately preserved surface GLP-1R compared with SRB106 (Figure 3C, D).We note that this difference was less marked than in SNAP-GLP-1R cells, which could reflect receptor species-related differences (as in Figure S2), other cell- We conclude from these experiments that differences in GLP-1R endocytosis and sustained cAMP signalling with G protein-biased agonists can occur at pharmacologically relevant concentrations.

| Complete attenuation of β-arrestin-2
recruitment is required for maximum antihyperglycaemic effects of biased glucagon-like peptide-1 receptor agonists The pharmacokinetics of semaglutide, SRB106 and SRB107 were compared in pigs, showing that both SRB peptides have similar half-life protraction, at approximately twice that of semaglutide (Figure 4A).This allowed us to test directly the effects of the different receptor pharmacology of SRB106 and SRB107 on physiological responses, with semaglutide also included as a therapeutically relevant comparator.
We first identified doses of each ligand with similar effects on appetite suppression: when assessed 72 h after a single dose, SRB106 and SRB107 at 3 nmol/kg were similarly anorectic to each other, and to 10 nmol/kg semaglutide (Figure 4B, C).Of note, at these doses, semaglutide led to markedly greater malaise acutely, as assessed by treatment-blinded behavioural observations and potentially indicative of nausea propensity (Figure 4D).We then tested for glycaemic effects via intraperitoneal (IP) glucose tolerance tests performed immediately at the time of dosing (d0) and after a 72-h delay (d3) (Figure 4E).Here, 10 nmol/kg semaglutide was highly effective acutely, but its anti-hyperglycaemic effect was lost by 72 h.At the sub-anorectic 1 nmol/kg dose, both SRB peptides showed little effect until the final time point of the acute IP glucose tolerance test, but interestingly, SRB107 was superior to SRB106 when assessed at 72 h.At 3 nmol/kg however, SRB106 and SRB107 were similar acutely, but only SRB107 was able to improve glucose tolerance at 72 h.SRB107 resulted in a greater plasma insulin response to IP glucose administration at the 72-h time point (Figure 4F), supporting a β-cell-mediated mechanism.There was no evidence of altered insulin sensitivity under the same conditions (Figure 4G).In diet-induced obese (DIO) mice, the same pharmacodynamic pattern was recapitulated, with 1.5 nmol/kg SRB107 but not SRB106 leading to improved glucose tolerance 72 h after dosing, and no clear difference in food intake or body weight effects between agonists at these doses (Figure 4H-J).
Finally, we performed a 28-day chronic administration study of SRB107 versus semaglutide in DIO mice, with both ligands dose-escalated from 1 to 5 nmol/kg over 19 days via daily subcutaneous injection.Notably, anorectic and weight loss effects of each agent were similar for the first 96 h but diverged thereafter, with SRB107 leading to greater food intake inhibition over time and an additional 10% body weight loss compared with semaglutide (Figure 4K, L).
These findings highlight how the anti-hyperglycaemic effects of SRB107 are longer lasting than both semaglutide and SRB106.While the former could be explained by a longer plasma half-life, this does not apply when comparing SRB106 and SRB107.We conclude therefore that the complete abolition of β-arrestin-2 recruitment with SRB107 probably allows it to retain signalling capacity and metabolic effectiveness over an extended period.

| Biased glucagon-like peptide-1 receptor agonist signalling is differentially affected by glucagonlike peptide-1 receptor coding variants
Finally, we investigated the potential for GLP1R coding variation to influence signalling properties of semaglutide and SRB107.Using  S1.
exome sequencing data from the gnomAD dataset 32 we identified single nucleotide coding GLP1R variants with an allele frequency of at least 1 Â 10 À3 in one or more of the genetic ancestral groups described (Figure 5A, B) and generated isogenic stable cell lines expressing each missense variant using the Flp-In system. 33Mutations from the signal peptide region (residues 1-23) were not included as the GLP-1R signal peptide in this system has been replaced with an N-terminal SNAP-tag.
We used the 'ΔΔlog(max/EC 50 )' approach to define the impact of each variant on acute cAMP signalling responses to semaglutide and SRB107, with the high affinity full agonist exendin-4 used as a reference ligand to reveal variant-specific differences in signal transduction. 34Note that this approach, rather than ascribing overall gain-or lossof-function to each variant, highlights differences in their ligand 'preference'.For example, compared with exendin-4, SRB107 is a low potency ligand at wild-type GLP-1R, but this loss of potency is somewhat mitigated by the D344E variant (Figure 5C).Across our panel of variants, we observed both disproportionate enhancement and disproportionate loss of test ligand responses; for example, the common G168S variant showed relatively enhanced responses to semaglutide and SRB107 compared with exendin-4, as did the low frequency A316T and D344E variants (Figure 5D).Note that in this analysis, statistical significance is inferred when the 95% confidence intervals shown do not cross zero.Re-analysis of the same data suggests that for A316T and D344E, this shift in ligand preference represents 'rescue' by semaglutide and SRB107 of a diminished exendin-4 response (Figure S4A).Contrasting with this, semaglutide and SRB107 responses with the V194I and A239T variants were disproportionately reduced, failing to rescue an already reduced response to exendin-4 (Figure 5D, Figure S4A).Overall, these differences were relatively modest, with the largest shifts in SRB107 preference observed with the D344E and A239T variants, amounting to a 5.5-fold selectivity difference (Figure 5D).Variant effects tended to be similar for SRB107 and semaglutide (Figure 5E).
We also performed experiments to determine the impact of more than one natural GLP1R variant being simultaneously present in the same receptor by generating 'double mutants' for the commonest GLP1R variants (Figure 5F, G, Figure S4B).Here, the effect of each additional mutation tended to imitate that of the same mutation present in isolation; for example, SRB107 signalling was enhanced by the presence of the A316T mutation irrespective of the presence of other coding variants.
These experiments show how some of the commoner GLP1R variants lead to ligand-specific changes in cAMP signalling.

| DISCUSSION
GLP-1R agonists are widely prescribed for T2D and obesity. 35The current generation of GLP-1R agonists used clinically, including semaglutide, show greater therapeutic efficacy with respect to weight loss than earlier iterations such as exenatide and liraglutide. 36,37Nevertheless, there remains an intense interest in the development of novel GLP-1R agonists for use either as monotherapy or as a component of multi-GPCR-targeting strategies. 38A more effective GLP-1R agonist might have a longer half-life to allow less frequent dosing, be more potent to allow for lower doses and manufacturing costs, or be more tolerable to increase the therapeutic window and achievable therapeutic efficacy.
Biased signalling is one potential route via which the pharmacological profile of GLP-1R agonists could be improved.Promising results have been obtained in several preclinical studies, [3][4][5][6][7][8] and the impressive performance in the phase 3 SURPASS-2 study of tirzepatide, 39 developed as a GLP-1R/GIPR co-agonist but also showing a GLP-1R signalling profile similar to several G protein-selective GLP-1R agonists, 10 places the concept of biased agonism centrestage.However, tirzepatide and other N-terminally modified G protein-biased GLP-1R agonists typically show reduced binding affinity, 9,11 probably because of the important role of the peptide N-terminus in orthosteric binding to the receptor. 40Here we have generated two novel ligands with similar acute cAMP signalling potency to semaglutide, but with reduced or absent recruitment of β-arrestin-2.Preserved acute signalling potency in the context of reduced desensitization tendency is predicted to be therapeutically advantageous.
β-arrestins have a canonical role in GPCR desensitization, but they may play a role in in orchestrating G protein-independent signalling events.GLP-1-regulated insulin release has been linked to β-arrestin-1 recruitment, 12 raising the possibility that abolishing this response could have undesirable consequences.However, β-cellspecific conditional knockout studies of β-arrestin-1 and -2 in mice showed no acute insulin secretory deficit in response to GLP-1R agonists, 41,42 although we recently found that loss of β-arrestin-2 may lead to reduced GLP-1R agonist-induced insulin secretion in female mice and DIO male mice, omitted from earlier studies. 43With this in mind, we aimed to test whether partial or complete loss of GLP-1R agonist-induced β-arrestin recruitment is optimal for enhanc- The rapid expansion of next-generation sequencing technology has highlighted extensive genomic coding variation in drug targets. 17rrent GLP-1R agonists are effective for many patients, but a significant minority fail to achieve treatment targets. 45Variation in the GLP1R coding sequence is one possible explanation, which may be the biology of biased GLP-1R agonism, namely: (a) is partial or complete loss of β-arrestin recruitment therapeutically preferable; (b) are GLP-1R internalization effects important at clinically relevant pharmacological concentrations; and (c) are biased agonist responses differentially modulated by naturally occurring GLP1R single residue missense GLP1R variants present in human populations.

1
Pharmacological characterization of novel glucagon-like peptide-1 receptor (GLP-1R) agonists.(A) Ligand amino acid sequences.(B) Mini-G s and β-arrestin-2 (βarr2) NanoBiT responses, and cyclic AMP (cAMP) in Flp-In T-REx-293-SNAP f -GLP-1R-SmBiT cells, n = 5, 30 min stimulation with GLP-1, semaglutide (Sema), SRB106 or SRB107, with three-parameter fits shown.(C) Relative signal transduction ratios (ΔLog τ/K A ) calculated from (B) relative to the mini-G s response for each ligand; statistical comparisons using mixed effects model with Tukey's test.(D) cADDis cAMP sensor signal in response to stepwise addition of increasing concentrations of each agonist in dispersed primary mouse islet cells (n = 5), INS-1832/3 cells (n = 6) or EndoC-βH5 cells (n = 5).Images show intensity changes in representative mouse islet cells at different agonist concentrations.(E) Relative signalling potency (ΔpEC 50 ) for each ligand, derived from data in (D) by determining concentration responses from area under the curve (AUC) during each time window and normalizing to GLP-1 for each system; statistical comparisons by two-way repeated measures ANOVA with Tukey's test.(F) Intracellular calcium responses in wild-type mouse islets treated with 100 nM agonist (GLP1RA) or vehicle (Veh) at 6 mM or 11 mM glucose (G6, G11), n = 5-6; statistical comparisons of AUC from indicated time window by two-way repeated measures ANOVA with Tukey's test.All data represented as mean ± SEM. *p < .05,**p < .01 by indicated test.

3 - 8
Whenexpressed T-REx 293 cells, INS-1832/3 cells or EndoC-βH327 β-cells, the SNAP-tagged GLP-1R appeared more membrane-resident after treatment with SRB107 than other ligands applied at 100 nM (Figure2A).A real-time assay 21 also showed notably slower clustering of SNAP-GLP-1R into endosomal puncta with 100 nM SRB107 compared with other ligands (Figure2B).Moreover, a reversible labelling assay, in which residual surface SNAP-GLP-1R labelling is stripped using the membrane-impermeant reducing agent Mesna, confirmed that less GLP-1R was internalized after 100 nM SRB107 than other ligands (Figure2C).Nevertheless, at very high agonist concentrations, more extensive GLP-1R internalization with SRB107 could be detected (Figure2D).To extend our analysis to untagged GLP-1R, we designed fluorescent conjugates of GLP-1, semaglutide, SRB106 and SRB107.For GLP-1 and semaglutide, tetramethylrhodamine (TMR) was appended at the C-terminus after a GK linker; for SRB106 and SRB106, TMR was inserted at position K12 (FigureS2A).cAMP signalling properties of each ligand in T-REx 293 cells stably expressing untagged human GLP-1R were preserved, except for a moderate reduction in potency of SRB107-TMR (FigureS2B).The expected pattern of ligand uptake was observed with both untagged or SNAP-tagged GLP-1R, with the latter showing strong colocalization of agonist and receptor (FigureS2C, D).
ences, we repeated these experiments in T-REx 293 cells transiently transfected with human or mouse GLP-1R (Figure S2G-J).

F
I G U R E 2 Legend on next page.
differences in endocytosis machinery, or assay considerations: residual binding of SRB107 at the cell surface may have reduced exendin-4-Cy5 binding during the labelling phase as both compete for the GLP-1R orthosteric site.As an additional corroboration, we measured TMR agonist uptake in INS-1832/3 cells after 12 h stimulation at 100 nM in the presence of 4% HSA, also showing reduced uptake of SRB107-TMR compared with SRB106-TMR (FigureS3A).Agonist-specific reductions in β-arrestin recruitment and GLP-1R internalization are expected to facilitate extended GLP-1R signalling, e.g.via avoidance of desensitization.To assess this in cells with endogenous GLP-1R expression, we measured the effect of 12 h of treatment with 100 nM of each agonist in the presence of 4% HSA on steady-state cAMP levels in both INS-1832/3 cells and intact mouse islets from CAMPER mice31 globally expressing the cAMP FRET sensor T Epac VV .In both cases, cAMP levels at steady state were increased with SRB107 compared with SRB106 and semaglutide (Figure3E, F).

[
Log(b/d) -Log(b/d)] -[Log(B/D) -Log(B/D)] WT D344E F I G U R E 5 Effect of GLP1R missense coding variants on signalling responses.(A) Population prevalence of indicated GLP1R missense variants in the gnomAD dataset, with breakdown according to recorded genetic ancestry.(B) Snake plot from gpcrdb highlighting the position of evaluated GLP1R variants.(C) Pooled cAMP response to exendin-4 (Ex4) and SRB107 in wild-type and D344E Flp-In T-REx-293-SNAP f -GLP-1R-SmBiT cells, n = 8, with mean ± SEM shown.Variant-specific potency ratios for SRB107 relative to Ex4 are shown on each graph, and the determination of ΔΔlog(E max /EC 50 ) is explained.(D) Semaglutide (Sema/SEMA) and SRB107 responses for single missense GLP1R variants, after normalization to exendin-4 for each variant using the ΔΔlog(E max /EC 50 ) approach, n = 8 for each variant.Data are represented as mean ± 95% confidence intervals, meaning that statistically significant effects are inferred when the confidence intervals do not cross zero.(E) Relationship between ΔΔlog(E max /EC 50 ) estimates for SRB107 and semaglutide for variants shown in (D).(F) As for (D) but for double missense GLP1R variants; the relevant single missense variant data are the same as in (D).(G) Heatmap representation of data from (F), showing how double missense variant responses tend to mimic those of the individual missense variants included.Data for semaglutide are on the bottom left, and data for SRB107 are at the top right.
ing insulin secretion using SRB106 and SRB107.It was clear that the greater avoidance of desensitization with SRB107 obviates any possible role for β-arrestin recruitment in regulating insulin secretion, both in vitro and in vivo.It seems plausible that studies in which β-arrestins are genetically depleted, even using inducible gene knockout strategies, can result in complex effects on islet function, which obfuscate clear roles in GLP-1R signal transduction.More rapid methods to inhibit or deplete β-arrestins, e.g.pharmacological inhibitors or rapid degradation approaches such as Trim-Away,44 are needed to corroborate results from biased agonist studies to resolve important questions regarding the role of β-arrestins.The mechanism for improved metabolic effects of G proteinbiased GLP-1R agonists is presumed to revolve around reduced desensitization and/or avoiding depletion of surface GLP-1R.We aimed to test whether clinically relevant pharmacological concentrations of acylated GLP-1R agonists, in the presence of physiological albumin levels, are sufficient to induce the extensive degree of GLP-1R internalization needed to diminish signalling responses purely via restricting surface receptor availability.While acknowledging that the in vitro systems used carry their own limitations, we conclude that measurable GLP-1R loss can occur at plasma concentrations that are readily achieved during at standard clinical doses, and that this is reduced with SRB107 when compared with agonists with greater inherent GLP-1R endocytic propensity.On the other hand, even with the full agonist semaglutide, GLP-1R disappearance under these conditions was far from complete, and it is unclear whether this supports a lack of agonist-mediated GLP-1R downregulation as a primary driver of the distinct biased agonist effects.If not via avoiding extensive GLP-1R downregulation, how could biased GLP-1R agonists achieve longer lasting biological effects?Another plausible explanation is that individual GLP-1Rs occupied by G protein-biased versus non-biased agonist molecules can avoid signal termination through reduced tendency to recruit β-arrestin, as suggested from differences in coupling to mini-G versus β-arrestin-2 from our initial pharmacological characterization.This advantage should apply irrespective of fractional receptor occupancy, i.e. across a full concentration range, although the concomitant reduced efficacy for G protein coupling also needs to be considered.