Spatiotemporal GLP-1 and GIP receptor signaling and trafficking/recycling dynamics induced by selected receptor mono- and dual-agonists

Objective We assessed the spatiotemporal GLP-1 and GIP receptor signaling, trafficking, and recycling dynamics of GIPR mono-agonists, GLP-1R mono-agonists including semaglutide, and GLP-1/GIP dual-agonists MAR709 and tirzepatide. Methods Receptor G protein recruitment and internalization/trafficking dynamics were assessed using bioluminescence resonance energy transfer (BRET)-based technology and live-cell HILO microscopy. Results Relative to native and acylated GLP-1 agonists, MAR709 and tirzepatide showed preserved maximal cAMP production despite partial Gαs recruitment paralleled by diminished ligand-induced receptor internalization at both target receptors. Despite MAR709's lower internalization rate, GLP-1R co-localization with Rab11-associated recycling endosomes was not different between MAR709 and GLP-1R specific mono-agonists. Conclusions Our data indicated that MAR709 and tirzepatide induce unique spatiotemporal GLP-1 and GIP receptor signaling, trafficking, and recycling dynamics relative to native peptides, semaglutide, and matched mono-agonist controls. These findings support the hypothesis that the structure of GLP-1/GIP dual-agonists confer a biased agonism that, in addition to its influence on intracellular signaling, uniquely modulates receptor trafficking.


INTRODUCTION
Glucagon-like peptide-1 (GLP-1) is a pleiotropic hormone with broad pharmacological potential due to its ability to improve body weight, food intake, and glucose metabolism [1]. However, active GLP-1, which is primarily GLP-1 (7e36 amide) and to a lower extent GLP-1 (7e37), is subject to rapid proteolytic degradation and fast renal elimination [2e5]. Long-acting analogs with biochemical modifications in the GLP-1 sequence have been designed to overcome these limitations and are in clinical use for treating type 2 diabetes [6,7]. Despite molecular enhancements in time action, dose-dependent adverse effects limit the maximal efficacy and overall therapeutic potential of GLP-1R mono-agonists [8]. Single chimeric molecules with dual agonism at the receptors for GLP-1 and glucose-dependent insulinotropic polypeptide (GIP) improve body weight and glucose handling with superior potency to GLP-1R mono-agonists in preclinical [9,10] and clinical studies [11]. While GLP-1/GIP dual-agonists have advanced to phase 3 clinical trials for treating obesity and diabetes, the 1 contribution of GIPR agonism to these applications is questionable. Mice with GIP receptor (GIPR) depletion are protected from dietinduced obesity [12], and patients with type 2 diabetes show an impaired insulinotropic response to GIP infusion [13]. Antibodies antagonizing GIPR improve body weight and glucose metabolism in obese rodents and non-human primates [14]. Recent hypotheses to reconcile these discrepancies include GIPR agonists acting as functional GIPR antagonists, or alternatively that specific ligands engage unique receptor signaling, trafficking, and/or recycling dynamics, commonly referred to as biased-agonism [15]. Biased agonism at the GLP-1R has been linked to differential cellular desensitization capacities via differences in receptor internalization and/or b-arrestin recruitment, as has been shown for Phe1-substituted exendin-4 [16]. In addition to the GLP-1R agonists exendin-4 and oxyntomodulin, both of which demonstrate bias toward b-arrestin recruitment [17], a/b amino acid modifications to the GLP-1 backbone sequence can also result in differential GLP-1R signaling [18]. Likewise, the GLP-1/GIP dual-agonist tirzepatide (LY3298176; Eli Lilly, Indianapolis, IN, USA) was recently reported to favor phosphorylation of ERK1/2 relative to barrestin and Ga s protein at both target receptors [19]. The aim of this study was to assess the spatiotemporal GLP-1 and GIP receptor signaling, trafficking, and recycling dynamics mediated by select GLP-1R and GIPR mono-and dual-agonists. MAR709 is characterized as a balanced GLP-1/GIP co-agonist and is acylated with a C16 fatty mono-acid, which allows for once-daily time action in humans [20,21]. Tirzepatide is characterized as an imbalanced GLP-1/ GIP co-agonist that favors GIPR potency and is acylated with a C20 fatty di-acid at position 20, which allows for once-weekly time action in humans [22]. Our results show that both dual-agonists, MAR709 and tirzepatide, act as partial effectors at GLP-1R for G protein recruitment, receptor internalization, b-arrestin recruitment, Rab5 þ /Rab7 þ receptor trafficking, and endosomal G-protein recruitment, while retaining full-agonist capacity for cAMP production. Interestingly, despite showing a reduced receptor internalization rate, MAR709 acts as a full effector for stimulating GLP-1R incorporation into Rab11 þ recycling endosomes. At the GIP receptor, dual-agonists similarly act as full agonists for cAMP, but lack the G protein recruitment partial agonism profile and display limited receptor internalization and trafficking properties. Our data support the hypothesis that biased agonism with unique receptor signaling and trafficking properties might be a potential basis for enhanced metabolic benefits of these GLP-1/GIP dualagonists.

Peptide synthesis
Semaglutide was provided by Novo Nordisk (Bagsvaerd, Denmark). All of the other peptides were prepared via standard automated Fmoc/tBu solid-phase peptide synthesis on Rink Amide ChemMatrix resin. An orthogonal protecting group strategy was used to incorporate the protraction moiety onto the appropriate lysine side chain. Following synthesis, crude compounds were cleaved from the resin with 95:2.5:2.5 trifluoroacetic acid/water/triisopropylsilane. The crude compounds were purified by reversed-phase high-performance liquid chromatography (RP-HPLC) on a Luna C8 (2) preparative column with a gradient of water/acetonitrile containing 0.1% trifluoroacetic acid, then lyophilized to produce the desired compounds as white powders. Compound identity was confirmed via RP-HPLC-mass spectrometry. hGLP-1 (7e36 amide) was purchased from Anaspec (Cat #: AS-22463, Fremont, CA, USA). hGIP (1e42) was purchased from Anaspec (Cat #: AS-61226-1, Fremont, CA, USA). Madison, WI, USA) or 1:500 NanoGlo (Cat #: N1110, Promega, Madison, WI, USA). BRET 1 measurements were taken every 30 s for 2 min at 37 C using a PHERAstar FS multi-mode microplate reader with 430e485 nm and 505e590 nm dual filters. Baseline measurements were taken after 5 min of incubation with coelenterazine-h or NanoGlo. The cells were then treated with a vehicle or the respective agonists. The resulting ratiometric BRET signal between the interacting fluorophore and lumiphore was normalized by subtracting the background ratio (505e590 nm emission over 430e485 nm) of the Original Article vehicle-treated wells with the matched agonist-treated wells producing a signal defined as the "ligand-induced BRET ratio" [25]. The temporal data of the vehicle-corrected agonist measurement was then normalized to the baseline reading of the same well. The first BRET reading following treatment with agonist/vehicle was the subsequent measurement after the zero time point. Positive or negative incremental areas under the curves (þiAUC/-iAUC) were calculated where noted. Each experiment was independently performed at least three times, with at least two technical replicates for each group.
2.5. G-protein recruitment assay Mini-G protein probes translocate to ligand-bound active receptors retaining their specificity (Wan et al., 2018). To measure the ligandinduced recruitment of the Ga s , Ga q , Ga i , and Ga 12/13 , 50 ng DNA of the respective NLuc-tagged mini-G plasmid was co-transfected with 500 ng DNA of GLP-1R GFP or GIPR-GFP per well of a 6-well plate.
2.6. cAMP assay CAMYEL, a cAMP sensor using YFP-Epac-RLuc [23] was utilized to quantify cAMP accumulation with the temporal resolution. Then, 500 ng of CAMYEL DNA was co-transfected with 500 ng of DNA of untagged GLP-1R or GIPR per well in a 6-well plate. The experiments were performed in the absence of 3-isobutyl-1-methylxanthine (IBMX).

GPCR internalization assay
A GPCR internalization assay was established by measuring the loss of baseline resonance energy transfer between an intracellular plasma membrane marker Venus-KRAS and hGLP-1R-RLUC8 or hGIPR-RLUC8 [26]. Then, 500 ng of Venus-KRAS DNA and 300 ng of the respective RLUC8-tagged GPCR DNA were used per well in a 6-well plate.
2.11. HILO microscopy HEK293T cells were seeded onto 24 mm coverslips (Cat #: 631e1584, VWR, Radnor, PA, USA) and transfected with 500 ng of GLP-1R-GFP or GIPR-GFP over 24 h. HILO image sequences were acquired with a custom-built TIRF microscope (Cairn Research) based on an Eclipse Ti2 (Nikon, Tokyo, Japan) equipped with an EMCCD camera (iXon Ultra, Andor), a 488 nm diode laser, a hardware Perfect Focus System, a TIRF iLas2 module, and a 100Â oil-immersion objective (NA 1.49, Nikon). Coverslips were mounted onto metal imaging chambers with a plastic seal and filled with imaging medium (HBSS supplemented with 10 mM of HEPES). The objective and samples were maintained at 37 C in a heated enclosure. Images were acquired on MetaMorph software (Molecular Devices) using a frame exposure of 50e200 ms with an image acquired before ligand stimulation and a subsequent image taken every 30 s thereafter, up to 20 min. All of the images were analyzed using ImageJ.
2.12. Data analysis Data are represented as means AE S.E.M. Each experiment was independently conducted at least three times, each with at least two technical replicates. E max values were normalized to GLP-1 (7e36 amide) or GIP (1e42). Dose responses were fitted using non-linear regression. pEC50 and EC50 values were calculated using GraphPad Prism 8.0 (GraphPad, San Diego, CA, USA). Statistical analyses were calculated in GraphPad 8.0 using one-way analysis of variance (ANOVA) and corrected with Tukey's or Bonferroni's multiple comparison test. Differences are considered significant with an adjusted p value < 0.05.

MAR709 and tirzepatide differed from GLP-1R and GIPR mono-agonists in G protein recruitment
Ligand-induced (1 mM) capacity for receptor G protein recruitment was assessed using bioluminescence resonance energy transfer (BRET)based technology in HEK293T cells transiently transfected with the respective GFP-tagged receptors and mini-G constructs. The molecules evaluated included the native ligands GLP-1 (7e36 amide) and GIP (1e42), semaglutide (Novo Nordisk, Copenhagen, Denmark), the GLP-1/GIP dual-agonists tirzepatide (Eli Lilly, Indianapolis, IN, USA) and MAR709 (Novo Nordisk, Copenhagen, Denmark), and two molecules (fatty acyl-GLP-1 and fatty acyl-GIP) that are derived from the MAR709 sequence but had been structurally modified via single-or doublepoint mutations to only activate either GLP-1R or GIPR ( Figure 1). A table including the external company identifiers, in-text abbreviations, and amino acid sequence structures of the agonists is available (Supplementary Table 1). In GLP-1R þ HEK293T cells, GLP-1 (7e36 amide) strongly recruited Ga s and to a lesser extent Ga q , with no meaningful recruitment of Ga i and Ga 12/13 (Figure 2AeD). GIP (1e42) and fatty acyl-GIP did not stimulate G protein recruitment in GLP-1R þ HEK293T cells, while semaglutide and fatty acyl-GLP-1 elicited comparable responses relative to GLP-1 (7e36 amide) (Figure 2AeD). Relative to the GLP-1 mono-agonists, both GLP-1/GIP dual-agonists showed a decreased ability to recruit Ga s and Ga q , however, MAR709 demonstrated a higher capacity to recruit Ga s and Ga q compared to tirzepatide (Figure 2A,B). The chimeric structures of MAR709 and tirzepatide did not additionally diversify the G-protein families recruited to the receptor as evidenced by a lack of Ga i and Ga 12/13 recruitment (Figure 2AeD). In GIPR þ HEK293T cells, native GIP (1e42) predominantly recruited Ga s without meaningful recruitment of Ga q , Ga i , and Ga 12/13 (Figure 2Ee H). As expected, GLP-1 (7e36 amide), fatty acyl-GLP-1 and semaglutide all showed negligible effects on G protein recruitment in the absence of GLP-1R (Figure 2EeH). Relative to native GIP (1e42), Ga s recruitment following treatment with fatty acyl-GIP and tirzepatide was comparable, but with MAR709 it slightly decreased ( Figure 2E).
In summary, the GLP-1 mono-agonists and GLP-1/GIP dual-agonists primarily initiated Ga s recruitment, and to a lesser extent Ga q , at the GLP-1R and GIPR. In relation to the GLP-1 mono-agonists, both GLP-1/ GIP dual-agonists showed decreased Ga s and Ga q recruitment in GLP-1R þ HEK293T cells. In GIPR þ cells, tirzepatide led to comparable recruitment of Ga s relative to native GIP while MAR709 showed a slight decrease in Ga s recruitment.
3.2. MAR709 and tirzepatide were partial agonists for Ga s recruitment at GLP-1R but full agonists for cAMP production We next assessed concentration-response dependence in ligandinduced Ga s recruitment and evaluated how this capacity translated to cAMP production. At all of the tested concentrations, semaglutide and fatty acyl-GLP-1 showed comparable Ga s recruitment relative to native GLP-1 (7e36 amide) in GLP-1R þ cells ( Figure 3A). In line with  iAUC (þiAUC) representation of vehicle and baseline-corrected 30 min response to each agonist is expressed as mean AE SEM. Bonferroni's test, *p < 0.05, **p < 0.005, and ***p < 0.0005 using one-way ANOVA vs GLP-1 (7e36 amide) or GIP (1e42). Three independent experiments were performed with at least two technical replicates per group.
Original Article 4 our previous data (Figure 2A), MAR709 and tirzepatide both acted as partial agonists at the GLP-1R, stimulating a respective 59% and 31% maximal Ga s recruitment (E max ) relative to GLP-1 (7e36 amide) ( Figure 3A and Table 1). This partial agonism was independent of the measurement time after drug exposure ( Figure 3B). Interestingly, despite a reduced Ga s recruitment E max , the dual-agonists did not differ in cAMP E max compared to the GLP-1 mono-agonists ( Figure 3C and Table 1). This was further validated with a CAMYEL sensor saturation assay (Supplementary Figure 4A-C), with ligand responses falling below the saturation limit of the sensor. Hence, despite partial agonism at the level of G protein recruitment to GLP-1R, the dualagonists remained full agonists when considering cAMP generation. In terms of potency, all of the agonists displayed similar cAMP pEC50 values except for tirzepatide, which was significantly decreased relative to GLP-1 (7e36 amide) ( Figure 3C and Table 1). In GIPR þ HEK293T cells, we observed a comparable potency and efficacy for Ga s recruitment upon treatment with fatty acyl-GIP and both dual-agonists relative to native GIP (1e42) ( Figure 3D and Table 1), which was independent of the measurement time after drug exposure ( Figure 3E). MAR709 exhibited a slightly reduced efficacy at the GIPR, stimulating 81% of the Ga s recruitment E max elicited by native GIP (1e42) ( Figure 3D and Table 1). For cAMP production, both fatty acyl-GIP and MAR709 displayed a significantly superior pEC50 than that of GIP (1e42), while tirzepatide exhibited a significant 3fold reduction in potency ( Figure 3F and Table 1). Collectively, MAR709 and tirzepatide displayed unique agonism properties at their target receptors, retaining full cAMP efficacy at both the GLP-1R and GIPR despite relatively lower GLP-1R-specific Ga s recruitment efficacy and a slightly reduced relative potency of tirzepatide for cAMP production at the GIP receptor.

MAR709 and tirzepatide showed decreased receptor internalization relative to GLP-1R and GIPR mono-agonists
We next assessed ligand-induced receptor internalization and the recruitment of b-arrestin and Ga q . In hGLP-1R-Rluc8 þ HEK293T cells, semaglutide and fatty acyl-GLP-1 showed similar receptor internalization dynamics relative to GLP-1 (7e36 amide) ( Figure 4A,B). However, both MAR709 and tirzepatide showed strikingly decreased receptor internalization compared to the tested GLP-1R mono-agonists ( Figure 4A,B). Relative to GLP-1 (7e36 amide), the maximal ligandinduced GLP-1R internalization (E max ) of MAR709 and tirzepatide was 51% and 13%, respectively ( Figure 4A,B and Table 1). Likewise, decreased internalization of GLP-1R was also observed upon treatment of hGLP-1R-Rluc8 þ Min6 cells with MAR709 and tirzepatide relative to GLP-1 (7e36 amide) and GLP-1 mono-agonists (Supplementary Figure 5A-C). No significant differences were observed in the pEC50 values of the tested ligands in HEK293T cells. Decreased receptor internalization mediated by MAR709 and tirzepatide was also confirmed using live cell HILO microscopy in HEK293T cells expressing GLP-1R-GFP ( Figure 4C). While treatment with GLP-1 (7e36 amide) and semaglutide initiated rapid internalization of GLP-1R-GFP, MAR709 and tirzepatide showed the persistent presence of the ligandereceptor complex at the plasma membrane with strikingly less trafficking into the cytosol ( Figure 4C). These data collectively demonstrated that MAR709 and tirzepatide differed from the GLP-1R mono-agonists in that they showed prolonged receptor presence at the cell surface and reduced receptor internalization.
GLP-1R recruitment of b-arrestin 1/2 (b-arr1/2) has been shown to influence receptor trafficking and enhance extracellular signalregulated kinase 1/2 (ERK1/2) signaling [29]. In GLP-1R þ HEK293T cells, semaglutide stimulated 67% and 78% of the b-arr1 and b-arr2  Table 1), while no measurable response for either b-arr1 or b-arr2 was seen with tirzepatide ( Figure 4D,E and Table 1). Table 1 e Maximal (E max ) drug effects and affinities at the GLP-1R or GIPR target receptors. Data were generated in HEK293T cells transiently transfected to express GLP-1R or GIPR. E max , pEC50, and EC50 values were generated from doseeresponse values fitted to sigmoidal curves using a three-parameter non-linear logistic regression. The E max is the maximal response elicited by an agonist and is expressed as % of the maximum response of GLP-1 (7e36 amide) or GIP (1e42). The EC50 is the molar concentration in which an agonist produced half of the maximal response. The pEC50 is the negative logarithm of the EC50. Values are given for Ga s recruitment, cAMP accumulation, receptor internalization, b-arrestin 1 / 2 , and Ga q recruitment at the GLP-1R and the GIPR. All of the values were derived from the iAUC of a temporal response for each concentration/agonist and are expressed as mean AE SEM from at least 3 independent experiments with at least two technical replicates per group. Statistical significance was determined using one-way ANOVA and corrected with Bonferroni's multiple comparisons test. */ # /yp < 0.05. * vs GLP-1 (7e36 amide) or GIP (1e42). # vs semaglutide. y vs fatty acyl-GLP-1 or fatty acyl-GIP. NA ¼ no agonism significantly different than zero observed at 1 mM stimulation. Bold red ¼ with significant non-zero agonism at 1 mM stimulation but incomplete curve fit, last value at 10 mM used. Original Article GLP-1R recruitment of Ga q has been proposed to regulate GLP-1R internalization via an ERK1/2 pathway [30]. In line with this data and our demonstration of decreased GLP-1R internalization upon treatment with MAR709 and tirzepatide (Figure 4AeC), we saw a less efficacious Ga q recruitment response to the GLP-1R upon treatment with MAR709 and tirzepatide, in which 48% and 17% of the GLP-1 (7e36 amide) E max was achieved, respectively ( Figure 4F and Table 1). In hGIPR þ HEK293T cells, we observed sustained receptor internalization induced by GIP (1e42) but no meaningful internalization following treatment with either fatty acyl-GIP, the GLP-1 mono-agonists, or the dual-agonists ( Figure 5A,B). In detail, MAR709 and tirzepatide stimulated 4% and 18% of the GIP (1e42) receptor internalization E max ( Figure 5A,B and Table 1). Reduced capacity of the dual-agonists for GIPR internalization was also confirmed visually through live-cell microscopy. Fifteen minutes after compound administration, GIP (1e42) showed a high dissolution of the GIPR-GFPdefined plasma membrane border with greater punctate structure formation in the cytosol, while neither MAR709 nor tirzepatide evoked a similar dynamic ( Figure 5C). Unlike b-arr2, b-arr1 has been shown to lack a functional role in GIPR internalization and trafficking [31]. Relative to GIP (1e42) at the maximal concentration of 10 mM, a 36% and 35% b-arr2 recruitment response was observed in cells treated with MAR709 or tirzepatide ( Figure 5D and Table 1). A true comparison between GIPR-b-arr2 agonist E max was not possible due to an incomplete curve fit for GIP (1e42). However, these data collectively suggested that reduced barrestin 2 recruitment by the dual-agonists may have had a functional correlation in the observed reduction in GIPR internalization or trafficking by these molecules. Relative to the Ga q recruitment E max for GIP (1e42), treatment with fatty acyl-GIP displayed a similar efficacy while MAR709 and tirzepatide exhibited 68% and 85% of the maximal response ( Figure 5E and Table 1). In summary, these data showed that MAR709 and tirzepatide both differed from the native peptides, semaglutide, and the PK-matched  (Figure 6AeC). Similar patterns were also observed when assessing total Ga s recruitment to GLP-1R þ Rab5 þ endosomes ( Supplementary Figure 2A-C). No difference in Rab5 co-localization was observed between GLP-1 (7e36 amide), semaglutide, and fatty acyl-GLP-1 (Figure 6AeC). In a hGLP-1R-Rluc8 þ min6 b cell model, tirzepatide likewise stimulated reduced co-localization of GLP-1R into Rab5 þ endosomes compared to GLP-1 (7e36 amide) and GLP-1 mono-agonists (Supplementary Figure 5D-F). Within HEK293T cells, co-localization of GLP-1R with Rab7 positive (late) endosomes was reduced, with MAR709 and tirzepatide stimulating 62% and 24% of the response of GLP-1 (7e36 amide) (Figure 6DeF). This pattern was replicated in ligand-induced Ga s recruitment to GLP-1R þ Rab7 þ endosomes (Supplementary Figure 2D-F). Notably, differences in GLP-1R co-localization with Rab11-positive recycling endosomes were insignificant between treatments of MAR709 and GLP-1 (7e36 amide), but treatment with tirzepatide decreased by 54% (Figure 6Ge  I). Despite substantial Ga s recruitment to Rab11 þ endosomes, endosomal Ga s recruitment by MAR709 was significantly reduced compared to GLP-1 (7e36 amide) (Supplementary Figure 2F-I).
Regarding the Min6 cell model, due to either a lack of BRET signals or and Ga q -Nluc recruitment (E). The þ iAUC representation of vehicle and baseline-corrected 20 min (GIPR internalization), 30 min (b-arrestin1/2 recruitment), or 60 min (Ga q recruitment) temporal response to each agonist is expressed as mean AE SEM. Three independent experiments were performed with at least two technical replicates per group. the requirement for improved detection sensitivity, replication of ligand-induced GLP-1R co-localization with Rab7-and Rab11-positive endosomes was not observable for any agonist (Supplementary Figure 5G-J). In HEK293T cells, the general agonist relationship between the AUC of GLP-1R endosomal co-localization and endosomal Gprotein recruitment was positively linear, in which greater endosomal trafficking was associated with greater Ga s recruitment to the endosomal sub-compartment (Supplementary Figure 6A-C). In summary, these data indicated that MAR709 not only induced less GLP-1R colocalization into early and late endosomes but also comparably incorporated GLP-1R into Rab11 þ recycling endosomes to that of GLP-1 (7e36 amide) and semaglutide in HEK293T cells. In GIPR þ HEK293T cells, GIPR co-localization into Rab5 þ endosomes was similar upon treatment with GIP (1e42), fatty acyl-GIP, and tirzepatide; however, MAR709 achieved approximately 66% of this response ( Figure 7A,B). This pattern was also seen in Ga s recruitment to Rab5 þ endosomes. No meaningful co-localization was detected with GIPR at either Rab7 or Rab11 (Figure 7CeF). The lack of receptor co-localization with Rab7 þ and Rab11 þ endosomes was similarly associated with a lack of endosomal Ga s recruitment (Supplementary Figure 3D-H). Discrepancies between GIPR Rab5 þ co-localization and the lack of GIP receptor internalization by the dual-agonists likely reflected methodological differences and/or lack of Rab5 þ early endosome scission from the plasma membrane.
Original Article capacity for cAMP, likely an advantageous effect of signal amplification systems. Similarly, both MAR709 and tirzepatide evidenced full agonism for cAMP at the GIPR, but only MAR709 displayed characteristics of partial agonism with a slight reduction in Ga s recruitment efficacy. These data together were in line with previously established reports [19]. Since MAR709 and tirzepatide showed 100% sequence homology at positions 1e12, the observed differences between MAR709 and tirzepatide apparently resulted from sequence substitutions at positions 13e27 of the peptides or from the size and location of fatty acylation. The aforementioned differences in total and endosomal Ga s recruitment may play a role in the endosomal sorting of the internalized receptor to Rab7 þ /lysosomal pathways [32]. GLP-1R internalization is primarily caveolin-1/dynamin dependent [33], mediated by Ga q signaling [30], and does not require but is influenced by b-arrestin [16,34]. GLP-1R internalization has been linked to the degree of cellular desensitization and insulin secretion in vitro [16,35]. Tirzepatide has previously been shown to elicit reduced GLP-1R internalization relative to native GLP-1 [36]. Whether this effect also holds true for other dual-agonists has yet to be demonstrated. Both dual-agonists evaluated herein retained a higher presence of GLP-1R at the plasma membrane relative to the tested GLP-1R mono-agonists and similarly displayed corresponding partial agonism for b-arrestin 1, b-arrestin 2, and Ga q recruitment to the GLP-1R. A Phe1 substitution within an exendin-4 sequence has previously been described to reduce GLP-1R internalization and b-arrestin recruitment [16]. In line with this, reduced internalization is also observed with a (phenolic) Tyr1 present in the MAR709 and tirzepatide amino acid sequences. Both dual-agonists showed minimal GIPR internalization relative to GIP (1e42). Yet, both GIP (1e42) and tirzepatide elicited equal GIPR incorporation into Rab5þ early endosomes. Reasons for the  A and B), Venus-Rab7 þ late endosomes (C and D), and Venus-Rab11 þ recycling endosomes (E and F). The þ iAUC representation of vehicle and baseline-corrected temporal response to each agonist is expressed as mean AE SEM. Bonferroni's test, *p < 0.05, **p < 0.005, and ***p < 0.0005 using one-way ANOVA vs GIP (1e42). Six independent experiments were performed with at least two technical replicates per group. discrepancy might have originated in the methodology of how internalization was assessed. Ligand-induced GLP-1R endosomal trafficking has not yet been fully elucidated. We showed that MAR709 did not differ from GLP-1 (7e36 amide) and semaglutide in terms of eliciting GLP-1R co-localization with Rab11þ recycling endosomes. Whether this was a consequence primarily of internalized receptor diverting into recycling pathways or whether increased Rab11 co-localization induced by MAR709 was supplemented with recruitment of GLP-1R from the biosynthetic pathway has yet to be established. Given the low rate of GLP-1R internalization and incorporation into Rab5þ and Rab7þ endosomes, MAR709's high capacity for Rab11þ co-localization and its biased signaling profile demonstrated unique spatiotemporal pharmacology at the GLP-1R that may facilitate potential attributes of cellular sensitization. A caveat to the receptor trafficking experiments was the limited potential for aberrant Venus-Rab localization into nonspecific endosomal compartments occurring from over-expression associated changes in Rab trafficking patterns. Additionally, transferability of these findings to physiologically relevant b cells was restricted to the min6 b cell model, and hence represents a limitation of this work. Despite favoring GIPR over the GLP-1R, tirzepatide showed comparable efficacy and potency relative to MAR709 at multiple signaling pathways connected to the GIPR, with the exception of cAMP pEC50 in which MAR709 exhibited higher potency. At the GLP-1R, MAR709 displayed higher Ga s /Ga q signaling, receptor internalization, and barrestin recruitment relative to tirzepatide despite still acting as a partial agonist in each of these categories. In addition, MAR709 elicited a disproportional incorporation of the GLP-1R into Rab11 þ recycling endosomes. Collectively, our data showed that MAR709 and tirzepatide differed from the tested receptor agonists in G protein recruitment, receptor internalization, and endosomal trafficking, which together supports the hypothesis that biased agonism of these molecules might contribute to their beneficial metabolic action profile.

AUTHOR CONTRIBUTIONS
A.N. and S.O'B. co-conceptualized the project, designed and performed the experiments, analyzed and interpreted the data, and wrote the manuscript. M.B., G.G., and M.K. conducted the experiments and analyzed and interpreted the data. P.K. performed peptide synthesis. A.Z., M.K., K.S., R.D.D., M.H.T., and B.F. co-conceptualized the project, interpreted the data, and revised the manuscript critically. D.C. and T.D.M. conceptualized and supervised the experiments, analyzed and interpreted the data, and wrote the manuscript.