Endogenous Lipid-GPR120 Signaling Modulates Pancreatic Islet Homeostasis to Different Extents

G protein-coupled receptors (GPCRs) sense diverse extracellular stimuli and play key roles in cell–cell communication. Several ligand-GPCR circuits have been discovered in pancreatic islets. For instance, the GLP-1–GLP-1R circuit in Langerhans islets regulates both insulin secretion and islet regeneration (1–3). Acetylcholine produced by human α-cells modulates insulin secretion via muscarinic acetylcholine receptors (4,5). Moreover, epigenetic regulation of somatostatin (SST) secretion by endogenous urocortin3 (UCN3) via CRHR2 serve as key mechanisms for islet homeostasis (6,7). However, the signaling circuits involved with fatty acid (FA) GPCRs in pancreatic islets are not fully understood.

GPR120 is one of the top 20 genes determined to be related to type 2 diabetes by systems genetic analysis (8). GPR120 deficiency in mice leads to obesity (9). In pancreatic islets, GPR120 is expressed mainly in α- and δ-cells, and activation of GPR120 in δ-cells is reported to inhibit SST release (10–14). Although GPR120 is the receptor for exogenous n-3 FAs and several endogenous lipids, such as palmitic acid-hydroxy stearic acid (PAHSA) (15–19), the characteristics of the ligand-GPR120 circuits in islets are unclear. Moreover, at least Gq, Gi, and β-arrestins are known to couple with GPR120 in a ligand-dependent manner, thus generating signaling diversity (10,15,20–23). Notably, attenuation of harmful effects by amplification of the particular downstream signals of a single GPCR may have new therapeutic potential for diseases (24–28), such as oliceridine, a recently U.S. Food and Drug Administration-approved pain killer. In this regard, it is important to determine whether endogenous GPR120 ligands are produced in pancreatic islets and, if so, to elucidate their identities and signaling biases. In the current study, we set out to delineate potential LCFA-GPR120 circuit within islets.

Sst-Cre mice (stock no. 013044), Ins2-Cre mice (stock no. 003573), db/+ mice (stock no. 000697), and green fluorescent protein (GFP)-reporter mice (stock no. 007676) with a C57BL/6 background were obtained from The Jackson Laboratory and were used in our previous studies (7,25,29). β-arrestin1^−/− and β-arrestin2^−/− mice on a C57BL/6 background were obtained from Dr R.J. Lefkowitz (Duke University, Durham, NC) and G. Pei (Tongji University, Shanghai, China) and used in our previous study (7,25,29). Gcg-Cre mice (stock no. C001052) with a C57BL/6 background were purchased from Cyagen (Suzhou, China). All mouse work was performed with the approval of the Institutional Animal Care and Use Committee of Peking University Health Science Center and Shandong University Cheeloo College of Medicine.

Human pancreatic islet samples were provided by the Department of General Surgery, Qilu Hospital of Shandong University, with informed written consent from patients and approval by the Clinical Research Ethics Committee of Shandong University (ECSBMSSDU2018-1-023). Human pancreatic islets came from 6 male donors undergoing pancreaticoduodenectomy, aged 45 ± 3, and their BMI was 25.19 ± 0.56 kg/m².

Mouse GPR120, β-arrestin1, β-arrestin2, and G protein subunits (Gαi, Gαq, Gβ1, and Gγ2) genes were subcloned into the pcDNA3.1 expression vectors. GPR120-YFP, GFP10-Gγ2 Gαq-RlucII, and Gαi-RlucII were created by fusion or insertion of different sequences to constructs. The human GPR120 mutation R270H was generated using the QuikChange Mutagenesis Kit (Stratagene). All of the constructs and mutations were verified by DNA sequencing.

Total RNA was extracted from isolated mouse islets or single cells derived from the islets using TRIzol reagent (Invitrogen). cDNA synthesis was performed using a commercial quantitative (q)RT-PCR kit (Toyobo, FSQ-101). qRT-PCR was performed on a LightCycler qPCR apparatus (Bio-Rad) using FastStart SYBR Green Master Mix (Roche). All primer sequences used for qRT-PCR assays are listed in Supplementary Table 1.

Mouse pancreatic islets were isolated from adult mice and cultured as previously described (30,31). Human islets were isolated from human pancreas obtained from the Department of General Surgery, Qilu Hospital of Shandong University, through digestion by collagenase P at 37°C for 30–50 min. After stop digestion and sedimentation, the islets were hand-picked for collection using a stereoscopic microscope and were cultured overnight in DMEM containing 5.56 mmol/L glucose (Biological Industries), 10% FBS (Gibco), and 0.1% penicillin/streptomycin (Biological Industries).

The hormone secretion measurement, including that of insulin and SST, was performed as previously described by our group (7,29,31,32). Stock free FAs were dissolved in DMSO and then emulsified in modified Krebs-Ringer bicarbonate buffer with an Autotune series high-intensity ultrasonic processor until homogenous (19,33).

Mice were fasted for 16 h before the glucose tolerance test. The plasma glucose levels were measured in tail blood by using a FreeStyle Lite Glucose Meter (Roche). After measurement of the baseline glucose level (0 min), the mice were intraperitoneally injected with 2 mg/g body wt glucose, and the glucose levels at 30, 60, 90, 120 min after injection were detected.

The β-arrestin recruitment and Gβγ-dissociation Bioluminescence Resonance Energy Transfer (BRET) assay were performed as previously described (25,34). The BRET signal was examined by using a Mithras LB 940 Multimode Microplate Reader (Berthold Technologies). The ratio of luminescence at 530 nm–to–485 nm and 510 nm–to–400 nm represented the β-arrestin recruitment and Gβγ-dissociation BRET signal intensity, respectively.

The GloSensor cAMP assay was performed as previously described (30,35). The cAMP signal was examined using a Multimode Plate Reader (EnVision, PerkinElmer).

The islets or cells were fixed and subjected to fluorescence microscopic analysis using an LSM 780 laser scanning confocal microscope system (ZEISS, Oberkochen, Germany), according to our previous publications (7,31). The recruitment of β-arrestin1 and β-arrestin2 to pancreatic islets by GPR120 was examined by coimmunoprecipitation experiments using a method similar to that in our previous publications (24,25,37). Antibodies used are listed in Supplementary Table 2.

Islets isolated from db/+ and db/db mice or the supernatant (100 μL) of islets isolated from wild-type mice after treatment were obtained for metabolomics analysis as previously described (38). The ΔOA or ΔLA concentration values were calculated by measuring the OA or LA concentration in the islet supernatant under 11.1 mmol/L glucose conditions and then subtracting the concentration in the islet supernatant under 2.8 mmol/L glucose conditions.

All results in this study are presented as the mean ± SEM. An unpaired two-tailed Student t test was used to compare two groups. One-way ANOVA, followed by the Dunnett test, was used for comparisons between multiple groups. Statistical analysis and graph generation were performed using GraphPad Prism 8 software. P values <0.05 were considered statistically significant.

Data sets generated in this study are available from the corresponding author upon reasonable request.

Preincubation of mouse pancreatic islets with 30 μmol/L AH7614, a selective GPR120 antagonist, significantly promoted SST secretion in response to 11.1 mmol/L glucose stimulation (Fig. 1B–D and Supplementary Fig. 1F). Simultaneously, glucose-stimulated insulin secretion was significantly decreased after AH7614 incubation (Fig. 1D). In contrast, glucose-stimulated glucagon secretion was not significantly changed (Fig. 1E). The effects of AH7614 on mouse islet hormone secretion were recapitulated by adenovirus-mediated knockdown of GPR120 expression in islets, with no effect on the expression of GRP40 and other relevant FA receptors (Fig. 1F–H and Supplementary Fig. 1H and I). Previous findings and our results both indicated that GPR120 is mostly enriched in mouse pancreatic δ-cells (10–13) (Fig. 1A and Supplementary Fig. 1A–E). Using mouse pancreatic islets derived from Sst-cre^+/− mice and infected with GPR120-shRNA^f/f lentivirus in vitro, we were able to specifically knock down GPR120 expression in mouse pancreatic δ-cells and examine the contribution of GPR120 residing in δ-cells to islet hormone secretion (Supplementary Fig. 2A–D). Importantly, compared with islets from Sst-cre^+/− littermates infected with the shNC^f/f (negative control shRNA) virus, Sst-cre^+/− GPR120-shRNA^f/f islets exhibited higher SST secretion and lower insulin secretion in response to high-glucose stimulation (Fig. 1I and J). However, glucose-induced glucagon secretion did not significantly differ (Fig. 1K).

We also characterized endogenous GPR120 activity in human pancreatic islets by application of AH7614, which recapitulated the phenotype identified in mouse islets by significantly enhancing SST secretion in response to 11.1 mmol/L glucose stimulation (Supplementary Fig. 1G). These data suggest that the activity of endogenous GPR120, especially that residing in pancreatic δ-cells, regulates hormone release and paracrine interactions in islets.

We collected the mouse islet supernatant under 2.8 mmol/L glucose and 11.1 mmol/L glucose conditions and then examined the effects of these supernatants on GPR120 activation using in vitro recruitment assays for arrestin, a pivotal transducer of GPCR signaling. In HEK293 cells overexpressing GPR120-GFP and Rluc–β-arrestin2, the 11.1 mmol/L glucose-treated islet supernatant promoted more β-arrestin2 recruitment to GPR120 than the low glucose–treated islet supernatant (Fig. 2A).

We then used lipidomics to analyze the islet supernatants, focusing on free FAs and PAHSA, which have been reported to be endogenous GPR120 ligands in other tissues (16). In particular, unsaturated FAs were investigated because saturated FAs were reported to be much weaker agonists than unsaturated FAs to induce GPR120 activation (15,19). PAHSA levels were lower, while OA (C18:1) and LA (C18:2) levels were higher, in the 11.1 mmol/L glucose supernatant than in the low-glucose supernatant (Fig. 2B and Supplementary Fig. 2E). Because PAHSA has been reported to be a potent GPR120 agonist (16), the decreased PAHSA level is unlikely to contribute to the increased GPR120 endogenous activity under high-glucose conditions. To further investigate whether OA and LA are potential GPR120 ligands in pancreatic islets that respond to high-glucose stimulation, we calculated the absolute values of ΔOA and ΔLA concentrations by comparing the islet supernatants from the 11.1 mmol/L glucose condition and 2.8 mmol/L glucose condition, and the results revealed concentrations of 2.26 ± 0.02 μmol/L and 5.18 ± 0.02 μmol/L, respectively (Fig. 2B). Our data indicated that GPR120 was able to sense ΔOA or ΔLA under these conditions by a β-arrestin2 recruitment assay (Fig. 2C). Moreover, we examined the effect of MK8245 and SC26196 on GPR120 activity induced by islet supernatant at the 11.1 mmol/L glucose condition (Fig. 2D and E). Whereas MK8245 is an OA synthetase (stearoyl CoA desaturase 1 [SCD1]) inhibitor (39), SC26196 is reported as an LA desaturase (FA desaturase 2 [FADS2]) inhibitor (40) (Supplementary Fig. 2F). Preincubation with MK8245 for 8 h reduced the corresponding GPR120 activity and OA level, and preincubation with SC26196 increased the corresponding GPR120 activity and LA level (Fig. 2D–G). Internalization of GPR120 and colocalization of GPR120 with β-arrestin2 in response to OA or LA treatment further supported that OA and LA are potential GPR120 ligands in islets (Fig. 2H). Considering the broad roles of SCD1 and FADS2, it did not exclude other FA metabolites might also participate in endogenous GPR120 signaling within islets. Collectively, these data supported that the increase in OA and LA concentrations in pancreatic islets might contribute to GPR120 activation in our assays. Further experiments using quantitative metabolomics, pharmacological characterization of every component of individual FAs, and knockout models of SCD1 and FADS2 may help to investigate the detailed roles of LA and FA in endogenous GPR120 signaling circuit within pancreatic isles.

G protein and β-arrestin are both common signal transducers downstream of GPCRs (24,25,30,37,41–46). Understanding the G protein or arrestin signaling biases of GPR120 in response to different endogenous ligands may facilitate the delineation of GPR120 functions in different physiological contexts (20,21). In general, the concentration-dependent curves suggested that OA and LA exhibited Gi activities similar to that of the reference endogenous GPR120 ligand 9-PAHSA (16,47) (Fig. 3D). Although LA showed stronger Gq activity than 9-PAHSA, both OA and LA recruited less arrestin than 9-PAHSA, the reference molecule. In particular, LA showed significantly stronger Gq and β-arrestin2 activity than OA (Fig. 3A–D and Supplementary Fig. 3A). We then used the biased β-value, which can be derived through an operational model (43), to quantify the relative activity of one pathway over another compared with that of the pathway induced by the reference agonist (41) (Fig. 3E). The β-value showed that LA was more biased toward β-arrestin2 and Gq than OA (Fig. 3E and Supplementary Fig. 3C). These data indicate that OA and LA show different bias properties in a model system.

Both OA and LA inhibited glucose-stimulated SST secretion and increased insulin and glucagon secretion (Fig. 4A–C). Notably, the same concentration of LA (200 μmol/L) exerted greater effects on promoting insulin and glucagon secretion and inhibiting SST secretion than OA (200 μmol/L) (Fig. 4A–C and Supplementary Fig. 3D). Importantly, the effects of OA and LA on SST and insulin secretion were significantly reversed by δ-cell–specific GPR120 knockdown (Fig. 4D–G). Although application of the selective GPR40 antagonist GW1100 showed no significant effects on SST secretion modulated by OA or LA, it partially decreased insulin secretion in response to OA or LA stimulation (Supplementary Fig. 3E and F). Therefore, both GPR120 and GPR40 participated in insulin secretion in response to OA or LA stimulation. These results demonstrate the important role of pancreatic δ-cell GPR120 in contributing to OA and LA functions to regulate SST and insulin secretion from islets (Fig. 4D and E).

We next examined the modulation of glucagon secretion by OA and LA via GPR120 in pancreatic δ-cells. GPR120 knockdown did not have significant effects on glucagon secretion, which is consistent with previous research (48) (Fig. 1H). Upon stimulation with OA or LA, the glucagon release level was significantly increased, which was not reversed by knockdown of GPR120 in δ-cells (Fig. 4C and F). We speculated that the increased glucagon secretion in response to OA or LA stimulation may be due to the expression of GPR120 in pancreatic α-cells. The involvement of GPR120 in pancreatic δ-cells with glucagon secretion could be complicated because glucagon secretion is modulated by both SST and insulin (14,49–52). Further investigations should be conducted to clarify the detailed underlying mechanism of this phenotype in the future.

Inflammation and oxidative stress are two important pathways involved in islet dysfunction and diabetes development (53,54). We observed that both OA and LA were able to reduce the expression of the inflammation-related genes Ccl2, Ccl5, Cxcl10, and Cxcl12 and stress response markers Cyba, Sod2, Fth1, and Nos2 induced by combined tumor necrosis factor (TNF)-α, interleukin (IL)-1β, and interferon (IFN)-γ treatment in islets isolated from wild-type mice (55,56) (Fig. 5A and B). Notably, under conditions of specific knockdown of GPR120 expression in δ-cells, the anti-inflammatory and antistress response effects of the two ligands were significantly weakened, suggesting that GPR120 mediated these effects (Fig. 5C and D and Supplementary Fig. 4A and B). In particular, LA showed greater anti-inflammatory effects than OA at 200 μmol/L (Fig. 5A and C). To further explore the mechanism of the anti-inflammatory effects of GPR120 signaling in δ-cells, we isolated primary δ-cells from islets of Sst-cre^+/− mice pretreated with lentiviruses expressing a control sequence or GPR120 shRNA^f/f. We found that the expression and secretion of inflammatory factors in primary δ-cells increased significantly in response to proinflammatory cytokines, which were downregulated by stimulation with OA or LA (Supplementary Fig. 4C and D). Knockdown of GPR120 in δ-cells reversed the effects of OA and LA, providing evidence to support that GPR120 plays key roles in the anti-inflammatory effects of LA and OA on δ-cells (Supplementary Fig. 4C–E). These results provide a better understanding of the function of pancreatic δ-cells.

We investigated the concentration-dependent effects of OA and LA on hormone secretion and anti-inflammation in pancreatic islets (Fig. 5E–J and Supplementary Fig. 5A–F). As the concentration of OA and LA increased, insulin secretion showed a dose-dependent increase using mouse islets (Fig. 5E and F and Supplementary Fig. 5A and B). The half-maximal effective concentration (EC50) values for insulin secretion in response to LA and OA were 55.97 ± 4.38 μmol/L and 343.7 ± 27.41 μmol/L, respectively, which are separated by one magnitude. The Emax of LA-induced insulin secretion is also ∼1.5-fold that of OA. In contrast, SST secretion decreased with increasing concentrations of LA and OA, with higher potency found for LA (EC50 = 55.4 ± 5.63 μmol/L) than OA (EC50 = 218.2 ± 28.65 μmol/L). These results indicated that LA showed higher potency or greater efficacy on insulin and SST secretion than OA.

However, in response to combined TNF-α, IL-1β, and IFN-γ treatment, a much smaller dose-dependent separation between OA and LA was observed in terms of their abilities to inhibit the expression of two inflammation-related genes, Ccl5 and Cxcl10 (Fig. 5G and H). The separation window for the potency between LA and OA in terms of the effects on Ccl5 and Cxcl10 expression was less than twofold, and their efficacies were comparable. Knockdown of GPR120 offset the anti-inflammatory effect, confirming the dependence on GPR120 for these effects (Supplementary Fig. 5C and D). We then used the operational model to derive the biased parameter β-value, which indicated that LA exhibited insulin/SST secretion functional bias relative to OA (Fig. 5I and J and Supplementary Fig. 5E and F). Taken together, these results demonstrate that OA and LA exhibit biased islet functions in a dose-dependent manner via interaction with GPR120. LA and OA showed a larger functional separation window for insulin/SST secretion compared with that for the effects on anti-inflammatory gene expression.

Notably, the difference of the biased factor in arrestin signaling between OA and LA was similar to their biased factor β in insulin/SST secretion (Fig. 3E and Fig. 5I and J). We therefore examined whether the functional differences between OA and LA are mediated by β-arrestin2. When the synthetic GPR120 agonist TUG891 was used as the control, both OA and LA were found to promote the direct interaction between native GPR120 and β-arrestin1 or β-arrestin2 in mouse pancreatic islets by coimmunoprecipitation assays (Fig. 6A and B and Supplementary Fig. 6A–E). Deletion of β-arrestin2 partly eliminated the differences between OA and LA in promoting insulin secretion and inhibiting SST secretion (Fig. 6C and D and Supplementary Fig. 6F and G). However, there was no significant difference between the islets derived from β-arrestin1^−/− mice or their wild-type littermates in terms of insulin secretion under the presence or absence of GPR120 agonist stimulation (Fig. 6E and F). Previous studies have shown that GPR120 internalization with β-arrestin2 mediates anti-inflammatory effects in RAW 264.7 cells (15). We next used β-arrestin2^−/− islets to explore whether the anti-inflammatory and antistress response effects of both GPR120 ligands were mediated by β-arrestin2. We observed that the addition of OA and LA did not obviously decrease the secretion of inflammatory factors or reactive oxygen species (ROS) and nitric oxide (NO) production after β-arrestin2 deletion in pancreatic islets, which is consistent with the mRNA examination data (Fig. 6G–I and Supplementary Fig. 6H and I). In islets derived from β-arrestin1^−/− mice or Gq knockdown by transduction of the shRNA-Gq^f/f lentivirus in islets isolated from Sst-Cre^+/− mice, the anti-inflammatory effects of OA and LA showed no significant difference compared with islets derived from wild-type littermates or islet transduction with control shRNA, respectively (Supplementary Figs. 7A–D, and 8A–D and G). Pentoxifylline (PTX) treatment showed no significant effects on anti-inflammatory functions, but could partially reverse the effects of OA or LA on SST secretion (Supplementary Fig. 8E and F). Therefore, β-arrestin2–mediated signaling might serve as the major contributor to the greater effect of LA on islet hormone and anti-inflammatory effects than OA at the same concentration (200 μmol/L).

Notably, the cellular context of islet cell types could alter the signaling bias, especially due to the differential expression of G protein or arrestin subtypes between different cell types. Difficulties remain in measuring the G protein or arrestin activity directly in vivo. Therefore, we examined the expression profiles of different G protein or arrestin subtypes in primary pancreatic islet δ-cells to provide preliminary insight into how the expression profiles of pancreatic islet δ-cells may affect the signaling bias of GPR120 properties. Our data indicated that β-arrestin2, Gs, G11, Gi2, and Gi3 were more relatively enriched in pancreatic δ-cells compared with the brain (Supplementary Fig. 3B). This observation is consistent with the contribution of β-arrestin2 signaling to GPR120 function in δ-cells.

Not only was GPR120 expression decreased, but the levels of the islet GPR120 ligands OA and LA were also significantly reduced in islets derived from db/db mice (Fig. 7A and B). These data suggested that the endogenous islet ligand-GPR120 circuits were impaired in the diabetic db/db models. In contrast, there was little difference in PAHSA content in islets from db/db mice compared with those from db/+ mice (Supplementary Fig. 9A). Treating the islets isolated from db/db mice with 200 μmol/L OA and 200 μmol/L LA increased GPR120 receptor expression and glucose-stimulated insulin secretion and reduced SST secretion, with a greater effect mediated by LA than by OA at the same concentration (Fig. 7C and D and Supplementary Fig. 9B). In addition, the administration of OA and LA ameliorated inflammation and oxidative stress in the islets derived from db/db mice (Fig. 7E–G and Supplementary Fig. 9C and D). Moreover, LA significantly ameliorated the glucose tolerance impairment in db/db mice, while OA showed no significant effects (Fig. 7H). In streptozotocin (STZ)-induced diabetic mice, we also found that OA and LA could restore insulin secretion and antagonize inflammation in isolated islets (Fig. 7I–K and Supplementary Fig. 9F and G).

The single nucleotide polymorphism p.R270H mutation of GPR120 is a deleterious nonsynonymous mutation and is reported to be associated with human obesity (9,57). Importantly, both β-arrestin recruitment and G protein dissociation in p.R270H of GPR120 was decreased compared with that of wild-type GPR120 in response to stimulation with OA or LA (Fig. 8A–I). The Gi activity of GPR120 p.R270H in response to LA was particularly impaired (Fig. 8F). Notably, only the EC50 of OA stimulated the activity of the Gi pathway of GPR120 p.R270H, and the EC50 of LA-stimulated activity of β-arrestin2 of GPR120 p.R270H was lower than 200 μmol/L. Notably, a previous study has suggested that the p.R270H variant led to increased fasting glucose levels (57). We therefore speculated that the p.R270H variant of GPR120 may cause impaired pancreatic islet δ-cell function, which contributes to the fasting glucose level and could be investigated in human samples or in GPR120 p.R270H knock-in mice in the future.

Paralleling our work, a recent GPR120 study (11) found that activation of GPR120 expressed in pancreatic δ-cells by synthetic agonists was able to modulate hormone secretion of islets. In the current study, we found that OA and LA may play a role as potential endogenous GPR120 ligands in response to high-glucose stimulation. Our measured OA and LA concentrations of 100 μL supernatant from 200 isolated islets were 4.2 ± 0.3 μmol/L (OA) and 3.6 ± 0.2 μmol/L (LA) under low-glucose conditions and 6.5 ± 0.2 μmol/L (OA) and 8.8 ± 0.2 μmol/L (LA) under high-glucose conditions. If we assumed that each islet had a 10 nL volume, the estimated intraislet concentrations of OA and LA could reach 322.9 μmol/L and 440.8 μmol/L, respectively. Therefore, we estimated that the intraislet concentrations of OA and LA could range from 3.6 μmol/L to 440.8 μmol/L. FAs in vivo are presented in various forms, including the aqueous-phase form that is bound to proteins, such as albumin, as well as the free fraction. Whereas lipidomics could provide useful information for the free fraction of FAs, the fraction of FAs bound with proteins is still difficult to quantify. Importantly, recent research reported a genetically encoded GPCR activation-based fluorescent sensor that enables rapid and specific detection of endogenous ligands, such as dompamine and cannabinoid, in vivo. Thus, this newly developed technology could facilitate the analysis of GPCR ligands in vivo (58–60). We are currently working with Yulong Li’s group to develop a genetic sensor that will enable the real-time detection of concentrations of free FAs in vivo. Notably, the metabolic pathways of LA and OA production and degradation are mediated by a panel of enzymes, including, but not limited to, FA synthase (FASN), SCD1, adipose triglyceride lipase (ATGL), and hormone-sensitive lipase (HSL) (61). In addition, the transportation of OA and LA contribute to extracellular concentrations of LA and OA, which requires the functions of specific transporters, such as FA translocase (FAT, CD36), FA transport protein (FATP), and the membrane FA-binding protein (FABP) families, as well as caveolins (62). Under high-glucose conditions for 1 h, the mRNA levels of the abovementioned proteins were not significantly changed. We therefore speculated that the enhanced extracellular OA and LA concentrations are related to the posttranslational modification-regulated functions of these proteins, which awaits further investigation.

Importantly, the two potentially endogenous ligands of GPR120 within islets, OA and LA, exhibited different signaling biases toward G protein subtypes and arrestin in a modeled HEK293 system. In HEK293 cells, LA was more biased toward the Gq and β-arrestin2 pathways than OA (Fig. 3B, C, and E). Because pancreatic δ-cells represent only a small proportion of islet cells, the measurement of second messengers or determination of receptor-arrestin interactions for biased characterization of GPCRs in pancreatic δ-cells is still very challenging and requires the development of new technology. In the current study, we used islets derived from β-arrestin1^−/− or β-arrestin2^−/− mice or specifically knocked-down Gq in pancreatic δ-cells to study the contribution of the functions of these effectors. We found that β-arrestin2 mediated SST/insulin secretion, the antistress response and the anti-inflammatory effects of several potentially endogenous GPR120 ligands in islets. The Gi signaling downstream of GPR120 may contribute to the SST secretion modulated by OA or LA. The combination of the application of GPR120-biased ligands, specific antagonists, and genetic knockdown all suggested that these phenotypes are very likely to be modulated directly by the GPR120–β-arrestin2 complex, which is similar to previous findings that the GPR120–β-arrestin2 complex mediated anti-inflammatory effects (15). However, our results do not exclude the possibility that other GPCR-arrestin interactions downstream of GPR120 may also be involved. Therefore, specific approaches, such as GPR120-arrestin chimeras or GPR120 ligands with exclusive bias properties, may help to address such questions in future research.

β-Arrestins are important downstream effectors that mediate a variety of GPCR functions (24–26,37,63–65). It has been reported that GPR120-β-arrestin2 signaling mediated the anti-inflammatory effects in macrophages by disrupting the TAB1-TAK1 complex and inhibiting the NLRP3 inflammasome (15,66). In adipocytes, GPR120 stimulation modulates peroxisome proliferator–activated receptor-γ phosphorylation via the association of β-arrestin2 with extracellular signal–regulated kinase (67). In addition, acceleration of wound healing by engagement of docosahexaenoic acid with GPR120 is dependent on β-arrestin2 signaling (68). In islets, it has been reported that β-arrestin1 is an important pancreatic δ-cell regulator in response to activation of the adrenergic system (32). β-Arrestin2 is reported to be required for β-cell mass plasticity and unlikely to be involved in the regulation of insulin secretion (69). However, another study showed reduced insulin secretion in β-arrestin2–deficient β-cells (70). We speculated that arrestin functions are highly dependent on the cellular context and specific agonist-receptor engagement because recent studies have suggested that receptor phospho-barcode and seven transmembrane bundle-mediated interactions are highly relevant to diverse arrestin functions (24,71–73). In the current study, we found that β-arrestin2 mediated the hormone secretion, anti-inflammation, and antistress response effects of OA and LA using β-arrestin2–deficient mice. These studies collectively suggested that activation of GPR120–β-arrestin2 signaling is generally beneficial for metabolism. The functions and phosphoproteomics downstream of each GPCR-arrestin complex are different. Further in-depth studies of GPR120–β-arrestin2 signaling in pancreatic islets could facilitate the development of therapeutics targeting GPR120.

Notably, recent studies have reported that the GPR120 agonist Metabolex 36 could decrease SST secretion of δ-cells in a PTX-sensitive manner (10). However, another study showed that PTX preincubation had no effects on SST secretion inhibited by the GPR120 agonist cpdA (11). PTX concentrations were similar in these two works, but the GPR120 agonists were different. It is possible that Metabolex 36 and cpdA have different abilities to activate the Gi and arrestin pathways. In particular, our data suggest that β-arrestin2 and Gi signaling contribute to changes in SST secretion downstream of GPR120 activation using pharmacological characterization combined with β-arrestin2^−/− mice. Future experiments characterizing the biased property between CpdA and Metabolex 36 and using Gi-specific knockout mouse models will help to clarify these discrepancies.

In addition, LA exhibited functional bias to modulate insulin/SST secretion compared with anti-inflammatory function, using OA as a reference ligand. There is an obvious functional separation window between OA and LA for insulin/SST secretion but not for the regulation of inflammatory gene expression. These results suggest that the different functions of GPR120 are connected to the bias of endogenous ligands.

Another significant finding was that endogenous GPR120 signaling circuits were downregulated within islets in the context of diabetes. In the human obese population, p.R270H nonsynonymous variants of GPR120 have been identified as a risk factor for obesity (9). In this study, we found that p.R270H in humans decreased its ability to sense the GPR120 ligands OA and LA. In particular, p.R270H was still able to respond to LA stimulation, despite requiring a higher concentration, suggesting therapeutic potential for LA or its synthetic analog. We further investigated whether the local GPR120 pathway in pancreatic islets is damaged in other diabetic models. We found that GPR120 circuits in islets were impaired in db/db diabetic model mice, with decreased GPR120 expression and decreased levels of the endogenous islet GPR120 ligands OA and LA. Treating islets isolated from db/db mice with LA or OA significantly improved insulin secretion and ameliorated inflammation as well as in the STZ diabetic model mice. Moreover, LA administration at a dose of 10 μg/g LA i.p. for 10 weeks was found to improve glucose metabolism in db/db mice. These results suggest that the endogenous GPR120 circuits are damaged in this diabetic model and that restoring the endogenous GPR120 pathways alleviates diabetic phenotypes in pancreatic islets.

In summary, we have revealed endogenous ligand-GPR120 circuits in pancreatic islets. The two identified potentially endogenous GPR120 ligands, OA and LA, exhibit different biases toward the Gq and β-arrestin2 pathways in modeled HEK293 cells, and β-arrestin2 signaling contributes to improved islet hormone secretion and islet homeostasis. The GPR120 circuits are downregulated in diabetic models, and restoring these circuits via exogenous lipid administration improves islet function. Our research revealed important endogenous FA-GPCR circuits that harbor different signaling biases that are connected to their functions in pancreatic islets.

Funding. Support was received from theNational Natural Science Foundation of China Grant (92057121 to X.Y., 32130055 to J.-P.S., and 31701230 to Z.Y.), the National Science Fund for Excellent Young Scholars Grant (81822008 to X.Y.), the National Key Basic Research Program of China Grant (2019YFA0904200 to J.-P.S., and 2018YFC1003600 to X.Y. and J.-P.S.), the Major Fundamental Research Program of Natural Science Foundation of Shandong Province, China (ZR2021ZD18 to X.Y. and ZR2020ZD39 to J.-P.S.), the National Science Fund for Distinguished Young Scholars Grant (81773704 to J.-P.S.), Innovative Research Team in University Grant (IRT_17R68 to X.Y.), and the Key Research Project of the Natural Science Foundation of Beijing, China (Z20J00124 to J.-P.S.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.-Q.D. contributed to identify the GPR120 agonists, OA and LA, and examine the function and signal transduction. Y.-Q.D., X.-Y.S., J.C., J.W., J.-P.S., and X.Y., participated in data analysis and interpretation. Y.-Q.D., X.-Y.S., J.C., J.W., J.-P.S., and X.Y. wrote the manuscript. Y.-Q.D., X.-Y.S., J.W., and W.P. isolated mice islets and performed ex vivo experiments. Y.-Q.D., X.-Y.S., and J.-Y.L. performed immunofluorescence studies. Y.-Q.D., J.-P.S., and X.Y. designed the islet culture and all other experiments. J.C. contributed to provide the db/db and db/+ mice and perform the RNAscope experiments. J.C. and L.-J.Z. tested the β-arrestins recruitment of p.R270H variants. J.W. contributed to provide the STZ-induced mice model. W.-T.A. contributed to provide the β-arrestin2^−/− mice. X.-N.T. and P.X. performed preparation of GPR120 protein. X.-N.T. and P.X. performed preparation of GPR120 protein. Y.-F.X. provided the human islets. Y.-L.J. and M.L. provided insightful ideas. Z.Y. guided the BRET assay of β-arrestins recruitment and Gβγ-dissociation. J.-P.S. and X.Y. conceived and initiated the study of the GPR120 in islet circuits, as well as the biased signaling of GPR120 in pancreatic islets. J.-P.S. and X.Y. and supervised the overall project design and execution. All of the authors saw and commented on the manuscript. J.-P.S. and X.Y. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.