Targeting GLP-1 receptor trafficking to improve agonist efficacy

Abstract

Glucagon-like peptide-1 receptor (GLP-1R) activation promotes insulin secretion from pancreatic beta cells, causes weight loss, and is an important pharmacological target in type 2 diabetes (T2D). Like other G protein-coupled receptors, the GLP-1R undergoes agonist-mediated endocytosis, but the functional and therapeutic consequences of modulating GLP-1R endocytic trafficking have not been clearly defined. Here, we investigate a series of biased GLP-1R agonists with variable propensities for GLP-1R internalization and recycling. Compared to a panel of FDA-approved GLP-1 mimetics, compounds that retain GLP-1R at the plasma membrane produce greater long-term insulin release, which is dependent on a reduction in β-arrestin recruitment and faster agonist dissociation rates. Such molecules elicit glycemic benefits in mice without concomitant increases in signs of nausea, a common side effect of GLP-1 therapies. Our study identifies a set of agents with specific GLP-1R trafficking profiles and the potential for greater efficacy and tolerability as T2D treatments.

Fig. 1

Insulin secretion is predicted by agonist-induced GLP-1R endocytosis. a Relationship between maximal agonist-induced SNAP-GLP-1R internalization in CHO-SNAP-GLP-1R cells (90 min, n = 5) and maximal prolonged insulin secretion in INS-1 832/3 cells (16 h, n = 6), expressed as insulin secretion index (ISI), i.e., fold increase vs. 11 mM glucose (G11) alone; full data in Supplementary Fig. 1. b As for a, but insulin secretion with 100 nM agonist in MIN6B1 cells, n = 7. c, d As for a, b, respectively, but over 1 h, n = 5. e Sequences in single letter amino acid code for exendin-4 (ex4), exendin-phe1 (ex-phe1), and exendin-asp3 (ex-asp3). f Net GLP-1R internalization (90 min) in CHO-SNAP-GLP-1R cells, full data in Supplementary Fig. 1, n = 5, four-parameter fit and statistical significance for E_max via one-way ANOVA with Dunnett’s test vs. exendin-4. g Prolonged insulin secretion in INS-1 832/3 cells, 16 h, n = 5, separate experiments to Supplementary Fig. 1, four-parameter fit and statistical significance for E_max via one-way ANOVA with Dunnett’s test vs. exendin-4. h Time-course insulin secretion in INS-1 832/3 cells, n = 5, two-way ANOVA with Dunnett’s test vs. exendin-4. i As for h, but in MIN6B1 cells, n = 5. Agonists applied at 100 nM, except where indicated. *p < 0.05, **p < 0.01, ***p < 0.001. Error bars indicate standard error of mean (SEM)

Fig. 2

GLP-1R agonist trafficking in MIN6B1-SNAP-GLP-1R cells. a Schematic for GLP-1R internalization and recycling measurements by FACS after labeling with cleavable SNAP-Surface probe. b Agonist-induced GLP-1R internalization in MIN6B1-SNAP-GLP-1R cells, n = 3, two-way randomized block ANOVA with Dunnett’s test vs. exendin-4. c GLP-1R recycling in MIN6B1-SNAP-GLP-1R cells 30 min after an initial 15 min agonist pulse to induce internalization; recycling measured in the presence of exendin(9-39) to block further endocytosis, n = 5 (exendin-4) or 3 (exendin-phe1 and -asp3), one-way ANOVA with Dunnett’s test vs. exendin-4. d Confocal images indicating GLP-1R internalization in MIN6B1-SNAP-GLP-1R cells, 30 min agonist incubation after SNAP-Surface-488 labeling, representative image from n = 3 experiments; scale bars, 8 μm. e Immunofluorescence showing increased co-localization of exendin-phe1-FITC vs. exendin-4-FITC and exendin-asp3-FITC with Rab11-positive recycling endosomes after 60 min agonist exposure, representative image from n = 2 experiments; scale bars, 4 μm. Individual red and green channels shown in Supplementary Fig. 12. f Representative electron micrographs showing subcellular localization of SNAP-GLP-1R (labeled with cleavable SNAP-Surface-biotin plus streptavidin-10 nm gold, arrows), 60 min agonist exposure; scale bars, 0.1 μm; larger area micrographs shown in Supplementary Fig. 5a. g Gold-particle quantification from f; n = 3 experiments, paired t-tests. PM plasma membrane, EE early endosome, MVB LM multivesicular body limiting membrane, MVB ILV multivesicular body intraluminal vesicle, Tub/RE tubular/recycling endosome, LE late endosome. h Immunoblots showing SNAP-GLP-1R (~73 kDa) levels in MIN6B1-SNAP-GLP-1R cells after 4 and 16 h agonist exposure, representative of n = 3 experiments. A smaller band possibly corresponding to deletion of GLP-1R C-terminal domain is detected under all conditions analyzed. i Confocal images demonstrating agonist-induced surface GLP-1R downregulation in MIN6B1-SNAP-GLP-1R cells; surface receptor labeled after 16 h agonist treatment; scale bar, 100 μm. j Quantification of experiments from i, five images analyzed per condition from n = 3 coverslips, mean cellular fluorescence indicated, one-way ANOVA with Dunnett’s test vs. exendin-4. k As for j, but quantified by FACS in separate experiments, results normalized to vehicle control, n = 4, one-way randomized block ANOVA with Dunnett’s test vs. exendin-4. Agonists applied at 100 nM. *p < 0.05, ***p < 0.001, by statistical test indicated above. Error bars indicate SEM

Fig. 3

Prolonged cAMP generation in beta cells. a cAMP measurement in INS-1 832/3 cells in response to continuous agonist exposure for indicated times. Incubations performed in the presence of 25 µM IBMX and the results expressed relative to IBMX-only response for each time-point, n = 4, one-way randomized block ANOVA comparing E_max with Dunnett’s test vs. exendin-4. b As in a, but with 100 nM agonist in MIN6B1 cells, n = 3, two-way randomized block ANOVA with Dunnett’s test vs. exendin-4. c cAMP accumulation dose response in INS-1 832/3 cells at the end of the 16 h agonist incubation; accumulation induced with 10 min addition of 500 µM IBMX, n = 6, one-way randomized block ANOVA comparing E_max with Dunnett’s test vs. exendin-4. d As in c, but 100 nM agonist in MIN6B1 cells, n = 5, one-way randomized block ANOVA with Dunnett’s test vs. exendin-4. e cAMP responses in INS-1 832/3 cells induced by the addition of 500 µM IBMX or 500 µM IBMX + 1 µM exendin-4 for the final 10 min after 16 h exposure to 1 µM agonist, expressed relative to response without agonist pretreatment, n = 4. f Response to 10 µM forskolin (FSK) in INS-1 832/3 cells pretreated with indicated agonist for 16 h, 10 min stimulation plus 500 µM IBMX, n = 5. g As for f, but with MIN6B1 cells, n = 5. h Homologous desensitization in INS-1 832/3 cells exposed to the indicated agonist for 24 h, washout, 1 h resensitization, and rechallenge ± GLP-1 100 nM, n = 5, one-way randomized block ANOVA with Dunnett’s test vs. exendin-4. i Cytosolic Ca²⁺ response to the indicated doses of GLP-1 in PathHunter CHO-GLP-1R cells exposed to 1 µM agonist for 90 min before a 30 min resensitization period, expressed as peak fold change from baseline reading, n = 5. *p < 0.05, **p < 0.01, ***p < 0.001 by statistical test defined in the text. Error bars indicate SEM

Fig. 4

β-arrestin-biased signaling reduces insulin secretion. a Agonist-induced cAMP, β-arrestin-1 (βarr1), and β-arrestin-2 (βarr2) responses in PathHunter CHO-GLP-1R cells, 10 min incubation, n = 4–6; the four-parameter logistic fit of averaged data shown. b Web of bias, depicting relative pathway preference for each agonist; data represent the inverse logarithm of normalized log (τ/K_A) values derived from a normalized to a reference agonist (exendin-4) and a reference pathway (cAMP); for further details, see Methods. Note that, β-arrestin-1 log (τ/K_A) values for exendin-phe1 could not be calculated due to absence of detectable response. c, d as for a, b but for 90 min incubation. e Confocal images of MIN6B1-SNAP-GLP-1R cells transiently expressing β-arrestin-2-GFP, labeled with SNAP-Surface-549, and treated with indicated agonist for 5 min before fixation, representative images from n = 2 experiments; scale bars, 10 μm. Individual red and green channels shown in Supplementary Fig. 12. f Relationship between biased signaling (Supplementary Fig. 6) in PathHunter CHO-GLP-1R cells and maximal prolonged insulin secretion (Supplementary Fig. 1) in INS-1 832/3 cells. Association quantified by linear regression. g Effect of dual β-arrestin silencing on prolonged (16 h) exendin-4-induced insulin secretion in INS-1 832/3 cells, n = 4, paired t-test comparing E_max. h As for g, but in MIN6B1 cells, n = 5, paired t-test. i As for g, but in EndoC-βH1 cells with stable knockdown of β-arrestin-1 and -2 by lentiviral transduction of shRNAs, n = 5, paired t-test. Agonists applied at 100 nM, except where indicated. *p < 0.05, by statistical test indicated above. Error bars indicate SEM

Fig. 5

Binding kinetics influence GLP-1R recycling. a Dissociation curve indicating FRET between FITC-agonist complexed with surface SNAP-GLP-1R, after inhibition of internalization, using NaN₃ and 2-deoxyglucose (Supplementary Fig. 9), 30 min agonist exposure, washout, and exendin(9-39) blockade, n = 4. Unmodified (non-FITC) agonist b residence time (1/k_off), c association rate constant (k_on), and d affinity, measured by TR-FRET in competition with exendin-4-FITC, with internalization inhibitors as above, and calculated using competitive kinetic method, n = 5, one-way randomized block ANOVA with Dunnett’s test vs. exendin-4. e Confocal fluorescence indicating co-localization of exendin-4-FITC or exendin-phe1-FITC with SNAP-GLP-1R (labeled with SNAP-Surface-549) after 60 min agonist exposure in MIN6B1-SNAP-GLP-1R cells, representative images from n = 2 experiments; scale bars, 8 μm. Individual red and green channels shown in Supplementary Fig. 12. f Schematic illustrating endosomal binding protocol. SA-Tb streptavidin-terbium cryptate. g Real-time FRET measurement of FITC-agonist complexed with internalized SNAP-GLP-1R after 30 min agonist exposure, washout, exendin(9-39) blockade, and cleavage of SNAP-biotin from surface SNAP-GLP-1R with MesNa, n = 5. Exendin-4 h residence time and i association rate constant ± 10 μM BETP, measured by TR-FRET in competition with exendin-4-FITC, n = 4, paired t-test. Exendin-4-induced j internalization (30 min), and k recycling (60 min) ± 3 μM BETP, n = 4, paired t-test. l Prolonged insulin secretion with exendin-4 ± 3 µM BETP in INS-1 832/3 cells, 16 h, n = 5, paired t-test for E_max assessed by four-parameter fit. Exendin-4 m cAMP, and n β-arrestin-2 responses, in PathHunter CHO-GLP-1R cells ± 3 μM BETP, n = 3. Agonists applied at 100 nM, except where indicated, and performed in CHO-SNAP-GLP-1R cells, except where indicated. *p < 0.05, ***p < 0.01, by statistical test indicated above. Error bars indicate SEM

Fig. 6

Comparison with licensed GLP-1R agonists. a Agonist residence time, measured by TR-FRET in competition with exendin-4-FITC in CHO-SNAP-GLP-1R cells, normalized to exendin-4-FITC response per assay, n = 5, one-way randomized block ANOVA with Dunnett’s test vs. exendin-4 (ex-phe1 not included in statistical analysis). Agonist-induced b cAMP, and c β-arrestin-2 responses in PathHunter CHO-GLP-1R cells, 90 min, n = 5. d Bias calculated from data in b, c, 95% CI shown. Agonist-induced e internalization (60 min), and f recycling (60 min, measured in presence of 10 μM exendin(9-39) to block rebinding) in CHO-SNAP-GLP-1R cells, normalized to GLP-1 response per assay, n = 5, one-way randomized block ANOVA with Dunnett’s test vs. exendin-4 (ex-phe1 not included in statistical analysis). Note that exendin-phe1 results, from a different set of experiments, are shown in a and d–f, for purposes of comparison, after normalization to a reference ligand on a per assay basis. g Principal component analysis taking into account agonist k_off, Δlog (τ/K_A) for cAMP and β-arrestin-2, internalization (60 min), and recycling (60 min). h Prolonged insulin secretion in INS-1 832/3 cells, 16 h, n = 6, one-way randomized block ANOVA with Dunnett’s test vs. exendin-phe1. Agonists applied at 100 nM, except where indicated. *p < 0.05, **p < 0.01, ***p < 0.001, by statistical test indicated above,. Except where indicated (bias plot), error bars indicate SEM

Fig. 7

Effects of exendin-phe1 in human islets. a Residual human islet surface GLP-1Rs labeled with exendin-4-FITC after overnight treatment with G11 ± agonist, representative image from n = 3 donors; scale bars, 10 μm. Individual red and green channels shown in Supplementary Fig. 12. Ca²⁺ responses in Fluo-2-loaded human islets to b acute stimulation with exendin-4 or exendin-phe1 at G11, or d to exendin-4 after overnight pretreatment with exendin-4 or exendin-phe1, n = 2 donors, 13–21 islets per condition analyzed. c, e Area under the curve (AUC), determined from the point of agonist addition (4 min) to the point of KCl addition (19 min), for traces depicted in b, d, respectively, relative to individual islet baselines, unpaired t-test. f cAMP response to 30 min of 100 nM GLP-1 + 500 µM IBMX in human islets treated overnight ± 100 nM agonist, expressed relative to vehicle pretreatment, n = 5 donors, paired t-test. g Prolonged insulin secretion in human islets, n = 11 donors, 16 h, paired t-test. h As for g but for 1 h stimulation, n = 8 donors, paired t-test. Agonists applied at 100 nM. *p < 0.05, **p < 0.01, by statistical test indicated above. Error bars indicate SEM

Fig. 8

Antidiabetic effects of exendin-phe1 in vivo. a Blood glucose during IPGTT (2 mg kg⁻¹ glucose) performed in HFHS mice at the indicated time-points after intraperitoneal (IP) injection of agonist (2.4 nmol kg⁻¹), n = 10/group except vehicle group at 4/8 h (n = 9), two-way repeat measures ANOVA with Tukey’s test, with significance vs. exendin-4 shown. b AUC determined from a, relative to baseline glucose at t = 0, the two-way ANOVA with Dunnett’s test vs. 0 h. c Plasma insulin before and 10 min after IP administration of glucose (2 g kg⁻¹) in HFHS mice, at indicated time-point after IP injection of agonist (2.4 nmol kg⁻¹), n = 10/group except vehicle 0 h (n = 8) and 4 h vehicle/exendin-4 (n = 9), two-way repeat measures ANOVA with Sidak’s test. d Plasma drug level at indicated time-points after IP injection of agonist (24 nmol kg⁻¹), n = 4 per group, two-way repeat measures ANOVA with Sidak’s test. e Cumulative food intake in fasted HFHS mice after IP injection of agonist (2.4 nmol kg⁻¹), n = 8/group. f Observed pica behavior in fasted lean mice treated with IP agonist (2.4 nmol kg⁻¹), n = 8/group, Mann–Whitney test comparing exendin-4 vs. exendin-phe1. g Relationship between agonist-induced GLP-1R internalization efficacy (Supplementary Fig. 1) and glucose lowering during IPGTT (Supplementary Fig. 10g), assessed as AUC relative to glucose at t = 0, relationship quantified by linear regression. *p < 0.05, **p < 0.01, ***p < 0.001, by statistical test indicated above. Error bars indicate SEM

Fig. 9

Effects of chronic treatment with exendin-phe1. a Fasting blood glucose in HFHS mice treated with continuous subcutaneous agonist or vehicle for 16 days, n = 10/group, one-way ANOVA with Tukey’s test. b Blood glucose during IPGTT (2 g kg⁻¹) performed after 14 days continuous agonist treatment, n = 10/group, two-way repeat measures ANOVA with Tukey’s test, with significance for exendin-phe1 vs. exendin-4 indicated. c AUC calculated from b, relative to baseline glucose at t = 0, one-way ANOVA with Tukey’s test. d Plasma drug level after 16 days treatment with indicated agonist (2.4 nmol kg⁻¹ day⁻¹) or vehicle in lean mice, n = 5/group, unpaired t-test. e Cumulative food intake, and f body weight change with continuous administration of agonist, n = 10/group. g Histologically determined steatohepatitis, quantified as nonalcoholic activity score (NAS), after 16 days agonist administration, n = 10/group, Kruskal–Wallis with Dunn’s test. Except where indicated (pharmacokinetic study), agonist administered at 0.24 nmol kg⁻¹ day⁻¹. *p < 0.05, **p < 0.01, **p < 0.001, by statistical test indicated above. Error bars indicate SEM