Impact of substitution registry on receptor-activation profiles of backbone-modified glucagon-like peptide-1 analogues

Abstract

Family B G protein-coupled receptors play important physiological roles and possess large extracellular domains (ECD) that aid in binding the long polypeptide hormones that are their natural agonists. We have previously shown that agonist analogues in which subsets of native α-amino acid residues are replaced with β-amino acid residues can retain activity while avoiding proteolytic degradation. This study focuses on eight new α/β analogues of glucagon-like peptide 1 (GLP-1) that each contain five α-to-β replacements in the C-terminal half of the peptide. This portion of GLP-1 is known to adopt an α-helical conformation and contact the ECD. All four registries ofαααβ backbone pattern were evaluated; previous work has shown that the αααβ patter supports adoption of an α-helix-like conformation. Two α-to-β replacement formats were employed, one involving β³ homologues of the native residues replaced and the other involving acyclic β residue. GLP-1R response was characterized in terms of stimulation of cAMP production and β-arrestin recruitment. Some of the backbone-modified GLP-1 analogues display biased agonism of the GLP-1R. This study helps to establish the scope of the α→β backbone modification strategy.

Graphical Abstract

Family B G protein-coupled receptors are activated by long polypeptide hormones, control an array of important physiological processes, and can exhibit pleiotropic signaling behavior. Here, we evaluated the effects of backbone modifications in the C-terminal region of Glucagon-like peptide-1; the region that interacts with the receptor’s extracellular domain. We observed varying degrees of activity and some agonists with functional selectivity.

G protein-coupled receptors (GPCRs) in family B are activated by long polypeptide hormones, and these receptors regulate a wide array of important physiological processes. Agonists of several family B GPCRs are used to treat human disease. Among these drugs are synthetic peptides containing unnatural components designed to optimize therapeutic performance.^[1–3] Unnatural modifications in clinical agents have so far been limited to side chains, but backbone-based modifications have been evaluated in the laboratory.^[4–14] We have focused on agonists in which a subset of natural α-amino acid residues is replaced by β-amino acid residues. This strategy is appealing because α→β replacements inhibit proteolytic degradation,^[15,16] which leads to very short in vivo lifetimes for natural peptide hormones. We have previously discovered that α→β modifications can modulate receptor selectivity or signalling profile relative to an all-α prototype.^[17–19]

The present study explores how varying the α→β replacement sites within the C-terminal region of glucagon-like peptide-1 (GLP-1) affects activation of the GLP-1 receptor (GLP-1R), an archetypal family B GPCR.^[20,21] This work is significant because current understanding of the functional consequences of α→β replacement in biologically active peptides is quite limited, and because this type of backbone modification is very easily implemented via standard solid-phase synthesis. The results reported here help to establish the scope of the backbone modification strategy.

Engagement of family B GPCRs by their polypeptide agonists has been viewed in terms of two components: binding of the agonist C-terminal region to the large extracellular domain (ECD) that is characteristic of this GPCR family, and binding of the agonist N-terminal region to the seven-helix transmembrane domain of the receptor.^[22] Co-crystal structures of agonists such as GLP-1 bound to the cognate ECD reveal that the C-terminal portion of the agonist usually adopts an α-helix in the ECD-bound state.^[23,24] Only recently has structural information become available for GLP-1 bound to full-length receptor; in this complex, α-helical secondary structure is maintained throughout the peptide agonist.^[25] Functional data for modified GLP-1R agonists, however, raise the possibility that receptor activation requires a non-helical conformation near the agonist N-terminus.^[26] The α→β substitution explored here are limited to the C-terminal portion of GLP-1 because the evidence for α-helicity in the receptor-bound state is strongest for this portion of the hormone.

We have previously reported that GLP-1 analogue A (Figure 1), which contains α→β substitutions at five positions in the C-terminal region and features an αααβ backbone pattern, can activate the GLP-1R in cell-based assays, as detected by stimulation of cAMP production.^[7] Crystal structures of other α/β-peptides have shown that the αααβ backbone pattern supports adoption of a conformation very similar to an α-helix.^[15] In addition to β residues, A contains two α-amino-isobutryric acid (Aib) residues, to minimize susceptibility to proteolysis in vivo.^[27] This α/β-peptide was indistinguishable from GLP-1 itself in its ability to induce insulin secretion from mouse pancreatic islets, and A displayed prolonged regulation of blood glucose concentrations in mice relative to GLP-1.^[7] All of the β residues in A contain a cyclic constraint that is known to stabilize an α-helix-like conformation.^[28,29] Initial evaluation of β residues that are homologous to the original α residues [e.g., Lys→β³-homolysine (β³-hLys), Ala→β³-hAla] as replacements in GLP-1 led to a drop in agonist potency in the cell-based assay.^[7] Although β³-homologues retain the natural side chain, they have enhanced backbone flexibility relative to cyclic β residues, which may increase the entropic cost of helix formation.^[30]

Figure 1:

Cartoon illustration the GLP-1 receptor (blue). The C-terminus of the peptide agonist (yellow) primarily interacts with the extracellular domain of the receptor.

The current study constitutes a systematic evaluation of periodic α→β substitution in the C-terminal portion of GLP-1. There are four ways to impose the αααβ pattern on a given α residue sequence, and all four have been examined here. One set of new GLP-1 analogues, 1a–4a (Figure 3), contains exclusively α→β³ replacements. These analogues retain all of the natural side chains but contain five additional CH₂ units in the backbone relative to GLP-1 itself; thus, α/β-peptides 1a–4a are isomers that differ only in the locations of the “extra” CH₂ units. The second set of analogues, 1b–4b, has substitutions at the same positions represented among 1a–4a, but all β residues are derived from (S,S)-trans-2-aminocyclopentanecarboxylic acid (ACPC). α/β-Peptides 1b–4b should therefore have greater helical propensity relative to 1a–4a, but the conformational stability is achieved at the expense of native side chains.

Figure 3:

Sequences of a/b peptide GLP-1 analogs 1a-4a and 1b-4b with αααß patterned substitution in the C-terminal region. The single letter code is used to indicate conventional proteinogenic amino acids, and colored circles indicate sites at which α-to-β replacements have been introduced.

Initial evaluations involved monitoring cAMP production stimulated by each of the new α/β-peptides in HEK293 cells transiently expressing GLP-1R along with stably expressed GloSensor™ protein for cAMP detection (Figure 3; Table 1).^[31] cAMP production is generally interpreted as an indication of G_s activation. The four analogues containing α→β³ replacements, 1a–4a, were very similar (p = 0.81) to one another in this assay; all were 50- to 60-fold less potent relative to GLP-1, but all reached the same maximum level of cAMP production as GLP-1. These results stand in contrast to observations with a different family B GPCR, the parathyroid hormone receptor-1 (PTHR1), which is activated by parathyroid hormone (PTH). The N-terminal fragment PTH(1–34), which is the osteoporosis drug teriparatide, is fully active as an agonist. A previous survey of PTH(1–34) analogues containing five α→β³ replacements in the C-terminal region, covering all four possible αααβ patterns, revealed that three of the four matched PTH(1–34) in potency for stimulating cAMP production, and the fourth analogue was only ~10-fold less potent.^[6] The data for 1a–4a reveal that the GLP-1R is more sensitive to α→β³ replacements in the C-terminal region of an agonist peptide than is the PTHR1, but the precise positioning of the backbone modifications has little effect on GLP-1R agonist activity. The disparate trends observed among comparably modified agonists of the GLP-1R and the PTHR1 might be related to functional differences between the receptors: the ECD of the GLP-1R is essential for receptor activation, but the ECD of PTHR1 is dispensable.^[32]

Table 1.

cAMP-production Potency and Maximum response for GLP-1 and α/β-peptides

cAMP Production
	pEC50	EC50 (nM)	% Max	EC50 rel.
GLP-1	10.4 ± 0.1	0.042	100	1
1a	8.76 ± 0.07	2.1	102 ± 4	50
2a	8.62 ± 0.08	2.4	102 ± 4	57
3a	8.57 ± 0.07	2.7	100 ± 4	64
4a	8.66 ± 0.1	2.2	103 ± 5	52
1b	8.64 ± 0.04	2.3	98 ± 2	55
2b	8.54 ± 0.03	2.9	99 ± 2	69
3b	< 7	>100	-	>1000
4b	9.44 ± 0.03	0.36	104 ± 1	9

EC₅₀ and maximal-response values for 3-parameter sigmoidal fits with data from ≥3 independent concentration-response experiments. EC_{50 rel.} indicates cAMP production potency relative to GLP-1 by the quotient (a/b-peptide EC₅₀)/(GLP-1EC₅₀). α/β-peptide 3b showed activity too weak to be measured in this assay.

GLP-1 analogues 1b–4b, containing the preorganized β residue ACPC, displayed substantial variations in agonist activity. α/β-Peptide 1b was 55-fold less potent in stimulating cAMP production relative to GLP-1. For this β residue registry, there was little difference between α→cyclic β replacements (as in 1b) and α→β³ replacements (as in 1a). There was a comparable parallel between 2a and 2b, although in this case α→cyclic β replacement (2b) caused a slightly larger potency decline relative to α→β³ replacement. α/β-Peptide 3b displayed the most profound potency loss among any of the new α/β-peptides presented here; no activation of the GLP-1R could be detected for 3b. In contrast, α/β-peptide 4b was the most potent among the eight GLP-1 analogues introduced in this study. The backbone pattern of 4a–b is the only αααβ registry for which replacing β³ residues with preorganized β residues (ACPC) enhanced agonist potency.

The inactivity of 3b may result from the absence of the Phe28 side chain. An alanine scan of GLP-1 revealed that Phe28→Ala was the most deleterious substitution to affinity, precipitating a >1000-fold loss of agonist potency.^[33] A co-crystal structure of GLP-1 with the GLP-1R ECD showed that the Phe28 side chain is buried at the peptide-protein interface.^[24], Meng et. al. reported a 10-fold and 5-fold loss of affinity and potency, respectively, upon substitution of Phe28 with hexafluoroleucine.^[34] The Ile29→Ala mutation caused a modest decline in agonist activity, but α/β-peptide 4b etained substantial potency despite the replacement of Ile29 with ACPC. This observation raises the possibility that the cyclopentyl side chain of ACPC can functionally mimic the isobutyl side chain of Ile in terms of interaction with the ECD.

GPCRs interact with diverse intracellular proteins to initiate multiple signaling pathways after agonist binding. In addition to G proteins, β-arrestins are important cytosolic partners that can influence duration and/or subcellular location of GPCR signaling as well as the identity of the signaling network.^[35] Proximity of a β-arrestin to a GPCR can be detected via bioluminescence resonance energy transfer (BRET), if each protein is appropriately modified. We used HEK393FT cells expressing fusion proteins GLP-1R-RLuc8 and either GFP²-β-arrestin-1or GFP²-β-arrestin-2(R393E, R395E) to assess recruitment of either β-arrestin-1 or β-arrestin-2 to the GLP-1R. The R393E, R395E variant of β-arrestin-2 minimizes signal loss arising from GLP-1R internalization.^[36] In both assays, G protein-coupled receptor kinase 5 (GRK5) was transiently transfected into the cells to enhance the maximum BRET signal.^[37] Phosphorylation of GLP-1R by GRK5 increases the binding of β-arrestins to the receptor. These two assays enabled us to determine how α→β replacements in the C-terminal portion of GLP-1 influenced β-arrestin recruitment (Figure 4; Table 2).

Figure 4:

GLP-1R activation as measured by cAMP concentration-response assay. (A) Measurements with α/β-peptides 1a-4a containing β³ residues. (B) Measurements with α/ß-peptides 1b-4b containing ACPC residues. Data points are the mean of ≥ 3 independent experiments. Error bars represent SEM.

Table 2.

Arrestin-recruitment Potency and Maximum response for GLP-1 and α/β-peptides

		β-Arrestin-1 Recruitment				β-Arrestin-2 Recruitment
	pEC50	EC50 (nM)	% Max	EC50 rel.	pEC50	EC50 (nM)	% Max	EC50 rel.
GLP-1	8.74 ± 0.1	3	100	1	8.72 ± 0.1	2.2	100	1
1a	6.59 ± 0.1	259	100a	86	7.51 ± 0.07	31	79 ± 3	14
2a	6.78 ± 0.1	169	60 ± 4	56	7.45 ± 0.1	35	66 ± 4	16
3a	6.96 ± 0.1	109	100a	36	7.39 ± 0.09	41	86 ± 4	19
4a	6.65 ± 0.1	232	100a	77	7.33 ± 0.2	47	71 ± 7	21
1b	7.15 ± 0.2	70	37 ± 5	23	8.18 ± 0.1	6.6	58 ± 3	3
2b	7.29 ± 0.2	51	20 ± 3	17	7.63 ± 0.1	23	56 ± 4	10
3b	< 6	>1000	-	>1000	< 6	>1000	-	>1000
4b	7.54 ± 0.1	29	65 ± 4	10	8.84 ± 0.1	3.6	69 ± 3	2

EC₅₀ and maximal-response values for 3-parameter sigmoidal fits with data from ≥3 independent concentration-response experiments. EC₅₀ rel. indicates cAMP production potency relative to GLP-1 by the quotient (α/β-peptide EC₅₀) / (GLP-1 EC₅₀). a/b-peptide 3b showed activity too weak to be measured in this assay.

^aFit maximum constrained to ≤ 100%

α/β-Peptides 1a–4a, containing flexible β³ residues, were all significantly less potent than GLP-1 itself in recruitment of both β-arrestin-1 and β-arrestin-2. Variation among EC₅₀ values was only approximately three-fold among these four α/β-peptides for β-arrestin-1, and the variation was even smaller among the four α/β-peptides for β-arrestin-2. Because of differences between the β-arrestin recruitment assay formats (two R→E modifications for β-arrestin-2 but not β-arrestin-1), it is not meaningful to compare the EC₅₀ values between these two assays. α/β-Peptides 1a, 3a and 4a appeared to approach the maximum level of β-arrestin-1 recruitment achieved by GLP-1, but the maximum obtained for 2a seemed to be significantly lower than that of GLP-1. None of these four α/β-peptides matched GLP-1 in terms of maximum β-arrestin-2 recruitment.

Among the α/β-peptides containing conformationally preorganized ACPC residues, 3b was unique in its inability to induce recruitment of either β-arrestin to the GLP-1R. Presumably this behavior stems from an inability of this α/β-peptide to bind to the receptor because the critical Phe28 side chain is missing, as discussed above. The other α/β-peptides containing ACPC residues, 1b, 2b and 4b, were less potent than GLP-1 in terms of recruiting either β-arrestin, and none of these agonists matched GLP-1 in terms of maximum level of β-arrestin recruitment. Among all eight α/β-peptides, 4b seemed most potent (lowest EC₅₀ values) in recruiting each of the β-arrestins, which parallels the superiority of 4b in terms of potency in the cAMP assay.

GPCR-mediated signaling is initiated when agonist binding affects receptor conformation in a way that is detected by intracellular protein partners. Different intracellular partners, such as G proteins and β-arrestins, appear to respond to different receptor states; therefore, the relative extents of signaling along distinct pathways can vary among different agonists.^[38,39] The signaling pattern evoked by the native agonist is generally taken as a reference point for a given receptor, and molecules that cause substantially different signaling patterns from the reference agonist, e.g., favoring one pathway relative to another in comparison with the reference agonist, are described as “biased.” We employed the Black-Leff operational model to quantify signaling bias among our new GLP-1R agonists.^[40,41]

Three of the eight α/β-peptides displayed statistically significant signal bias (Table 3). The strongest effect was observed for 1b, which was biased toward β-arrestin-2 recruitment vs. cAMP production relative to GLP-1. This finding is consistent with a previous study employing the same cell-based assays in which we observed that α/β-peptide B, the analogue of A that contains only the α→β replacements, was biased toward β-arrestin-2 recruitment vs. cAMP production relative to GLP-1.^[17] B and 1b feature the same β residue locations; these molecules differ only at the β residue closest to the C-terminus, which contains a cyclopentane ring in 1b but a pyrrolidine ring in B. A smaller bias toward β-arrestin-2 vs. cAMP production was observed for α/β-peptide 2b. α/β-Peptide 2a was modestly biased toward cAMP production relative to β-arrestin-1 recruitment. The most potent α/β-peptide, 4b, displayed balanced signaling behavior, comparable to that of GLP-1 itself.

Table 3.

Bias factors (ΔΔLog(t/K_A)) Calculated for GLP-1 and α/β-peptides

Bias Factors: (ΔΔLog(τ/KA)
	β-arr1 vs. cAMP	β-arr2 vs. cAMP
GLP-1	0 ± 0.2	0 ± 0.2
1a	−0.4 ± 0.2	0.3 ± 0.2
2a	−0.6 ± 0.2*	0.2 ± 0.2
3a	−0.1 ± 0.2	0.4 ± 0.2
4a	−0.3 ± 0.2	0.1 ± 0.2
1b	0.1 ± 0.2	0.9 ± 0.1*
2b	0.04 ± 0.3	0.4 ± 0.1*
3b	-	-
4b	0.1 ± 0.2	0.3 ± 0.2

Bias factors (ΔΔLog(τ/K_A)) calculated in terms of β-arrestin-1recruitment relative to cAMP production, and β-arrestin-2 recruitment relative to cAMP production. Positive values indicate bias towards arrestin-recruitment and negative values indicate bias towards cAMP-production.

^*indicates P < 0.05 compared to GLP-1 by one-way analysis of variance with Dunnent’s test.

We have evaluated a set of GLP-1 analogues containing periodic backbone modifications (α→β replacements) at five sites in the C-terminal region. The αααβ backbone pattern covers >50% of the length of each polypeptide; all four possible αααβ registries have been examined. Two different types of α→β modification have been studied, involving β³ residues (which retain natural side chains but introduce flexibility) or cyclic β residues (which are preorganized to favor an α-helix-like conformation but lack the side chain). Each of the α/β analogues of GLP-1 displays a decline in receptor activation potency relative to GLP-1 itself, as detected by three different measures of GLP-1R function: stimulation of cAMP production (an indirect indication of G_s activation), β-arrestin-1 recruitment and β-arrestin-2 recruitment. However, variations in the locations and identities of the β residues lead to substantial variations in signaling outcomes.

None of the α/β analogues of GLP-1 reported here matches the potency of the natural hormone, and we speculate that achieving this goal will require a departure from introducing α→β modifications in simple patterns. Such departures could involve reverting one or more sites (β→α) in an analogue with periodic substitution, or maintaining the number of α→β substitutions but deviating from the αααβ pattern. The most active α/β agonist reported here, 4b, is within 10-fold of GLP-1 potency in each of the three assays and therefore represents a good starting point for such efforts. Several of the α/β analogues display significant cAMP vs. β-arrestin signaling bias relative to GLP-1, which is intriguing given that all of the analogues are identical in the N-terminal region, the portion that is most intimately engaged by the transmembrane domain of the receptor. Overall, these results provide a foundation for future exploration of α→β replacements in GLP-1, including extension toward the N-terminus, in pursuit of potent GLP-1R agonists with maximum backbone modification, which should provide the most effective resistance to proteolysis.

Experimental Section

Full experimental details including list of instrumentation, list of materials, peptide purification, peptide characterization and, cell-culture can be found in the supporting information. Peptide Synthesis: Peptides were synthesized by standard solid-phase methods using NovaPEG rink amide resin.. cAMP-Production: Protocol was adapted from Hager et al. and Binkowski et al. HEK293 cells stably expressing the Glosensor-22F (Promega) luminescent cAMP-sensing protein were grown to confluence and then plated 1:3 onto a 10 cm tissue-treated dish with 10 mL of appropriate culture medium (vide supra) without penicillin/streptomycin. The cells were then incubated at 37°C with 5% CO₂ overnight. After the overnight incubation, the medium was aspirated, 4.5mL McCoy’s 5A modified medium with 10% FBS was added, the cells were incubated at 37°C with 5% CO₂. During this incubation, 10 μg of GLP-1R plasmid and 32 μL FuGENE HD transfection reagent was added to 1 mL of opti-MEM. After 20 min, 4.5 mL of DMEM with 10% FBS was added to the cells, and 1 mL of the transfection mixture was gently pipetted onto the medium. The cells were then returned for incubator overnight. The next day, cells were washed with DPBS, harvested with 0.05% Trypsin-EDTA, and resuspended into 4 mL appropriate medium without penicillin/streptomycin. The cell suspension was diluted to approximately 300,000 cells/mL in medium, and 100 μL of cell suspension was pipetted into each well (providing ~30,000 cells/well) of a clear-bottom, white-walled, 96-well plate. The cells in the 96-well plate were allowed to incubate at 37°C with 5% CO₂ overnight. After 24 h, the medium was removed by inverting and gently flicking the plate. DPBS with D-luciferin (500 μM) was quickly added to the plate (90 μL/well). The cells were allowed to incubate for approximately 20 min at room temperature before addition of peptide (as 10 μL/well diluted in DPBS). We found it to be important to change pipette tips between serial dilutions for reproducible results. The plate was then transferred to a BioTek Synergy 2 plate reader with no optical filter (“hole”), 1 mm vertical probe offset, and read with a sensitivity value of 200. Curves were generated from luminescence values observed between 10 and 20 min.

Experiments were conducted with n ≥ 3 biological replicates. Reported EC₅₀ and %Max values were a result of normalizing, averaging, and then fitting data to three-parameter sigmoidal curves in GraphPad Prism 6. The bottom of the curves was constrained to 0%. Normalization was performed with 100% representing the top of GLP-1’s curve for individual experiments and 0% representing the luminescence value in absence of peptide.

Arrestin-Recruitment: Protocol was adapted from Hager et al.^[17] and Jorgensen et al.^[36] HEK293FT cells were grown to confluence and then plated 1:3 onto a 10 cm tissue-treated dish with 10 mL of appropriate culture medium (vide supra). The cells were then incubated at 37°C with 5% CO₂ overnight. After 24 h, a transfection mixture was made with 1:1 polyethylenimine (PEI, 1 mg/mL in water, pH 7.0):DNA in 1 mL opti-MEM. Either GFP²-ß-arrestin-1 (14 μg) or GFP²-ß-arrestin-2(R393E, R395E) (14 μg) along with GRK5 (250 ng) and GLP-1R-Rluc8 (250 ng for ß-arrestin-1 experiments or 130 ng for ß-arrestin2 experiments) was added to the PEI/Opti-MEM mixture, and the resulting transfection mixture was incubated for 20 min at room temperature. The cell medium was then aspirated, replaced with 4.5 mL of DMEM without FBS, and the transfection mixture was gently pipetted into the medium. Six hours after transfection (with incubation at 37°C with 5% CO₂), 4.5 mL of DMEM supplemented with 20% FBS was added to the dish. Twenty-four hours after transfection (with incubation at 37°C with 5% CO₂), cells were washed with DPBS, harvested with 0.05% Trypsin-EDTA, and resuspended into 4 mL appropriate medium without penicillin/streptomycin. The cell suspension was diluted to approximately 1,000,000 cells/mL in medium, and 100 μL of cell suspension was pipetted to each well (providing ~100,000 cells/well) of a white-bottom, white-walled, 96-well plate. The cells in the 96-well plate were incubated at 37°C with 5% CO₂ overnight.

Twenty-four hours after adding cells to the 96-well plate, the medium was removed by pipette, the cells were washed twice with DPBS (with glucose, 100 μL/well), and 100μL of DPBS (with glucose) was added to each well. The cells were incubated at 37°C with 5% CO₂ for 45 min to1 h before addition of peptide (as 10 μL dilutions in DPBS). After addition of peptides, the cells were incubated at room temperature 20 min before addition of Rluc8 substrate, DeepBlueC (10 μL/well of 60 μM DeepBlueC in 2:1 DPBS:ethanol). The 96-well plate was transferred to a BioTek Synergy 2 plate reader with 400 nm (20 nm bandwidth) and 528 nm (30 nm bandwidth) optical filters, 1 mm vertical probe offset, 1-second integration time and read at maximum (200) sensitivity. Concentration-response curves were generated with I_528nm/I_400nm values taken between 15 and 45min after initial read, as signal variability was found to be relatively high at earlier timepoints.

Experiments were repeated n ≥ 3 biological replicates. Each experiment consists of at least 7 different concentrations of peptide, with solutions prepared via serial dilution of a stock solution of each peptide prepared from 1 mM DMSO stock. Reported EC₅₀ and %Max values were a result of normalizing, averaging, and fitting data to three-parameter sigmoidal curves in GraphPad Prism 6. The bottom of the curves was constrained to 0%. Normalization was performed with 100% representing the top of GLP-1’s curve for individual experiments and 0% representing the bottom value if the curves were fit with raw data and constrained to have a shared minimum.

Figure 2:

Sequences of GLP-1 and previously reported α/β-GLP-1 analogs. The single letter code is used to indicate conventional proteinogenic amino acids, and colored circles indicate sites at which α-to-β or Aib replacements have been introduced.

Figure 5:

GLP-1R activation as measured by a BRET-based arrestin recruitment assay. (A) β−arrestin-1 recruitment measurements with α/ß-peptides 1a-4a (B) β−arrestin-2 (R393E, R395E) recruitment, measurements with α/β-peptides 1a-4a. (C) β−arrestin-1 recruitment measurements with α/β-peptides 1b-4b (D) β−arrestin-2 (R393E, R395E) recruitment measurements with α/β-peptides 1b-4b. Data points are the mean of ≥ 3 independent experiments. Error bars represent SEM.