A New Pathway for Glucose-dependent Insulinotropic Polypeptide (GIP) Receptor Signaling

The hormone glucose-dependent insulinotropic polypeptide (GIP) is an important regulator of insulin secretion. GIP has been shown to increase adenylyl cyclase activity, elevate intracellular Ca2+ levels, and stimulate a mitogen-activated protein kinase pathway in the pancreatic β-cell. In the current study we demonstrate a role for arachidonic acid in GIP-mediated signal transduction. Static incubations revealed that both GIP (100 nm) and ATP (5 μm) significantly increased [3H]arachidonic acid ([3H]AA) efflux from transfected Chinese hamster ovary K1 cells expressing the GIP receptor (basal, 128 ± 11 cpm/well; GIP, 212 ± 32 cpm/well; ATP, 263 ± 35 cpm/well;n = 4; p < 0.05). In addition, GIP receptors were shown for the first time to be capable of functionally coupling to AA production through Gβγ dimers in Chinese hamster ovary K1 cells. In a β-cell model (βTC-3), GIP was found to elicit [3H]AA release, independent of glucose, in a concentration-dependent manner (EC50 value of 1.4 ± 0.62 nm; n = 3). Although GIP did not potentiate insulin release under extracellular Ca2+-free conditions, it was still capable of elevating intracellular cAMP and stimulating [3H]AA release. Our data suggest that cAMP is the proximal signaling intermediate responsible for GIP-stimulated AA release. Finally, stimulation of GIP-mediated AA production was shown to be mediated via a Ca2+-independent phospholipase A2. Arachidonic acid is therefore a new component of GIP-mediated signal transduction in the β-cell.

Glucose-dependent insulinotropic polypeptide (GIP, or gastric inhibitory polypeptide) 1 is a 42-amino acid polypeptide hormone synthesized by mucosal K cells of the duodenum and jejunum and released into the circulation in response to nutrient ingestion (1)(2)(3)(4). GIP and glucagon-like peptide-1 (GLP-1) are thought to be the major hormones (incretins) that constitute the endocrine component of the enteroinsular axis in humans and are responsible for at least 50% of postprandial insulin secretion (5). In non-insulin-dependent diabetes mellitus (type 2 diabetes mellitus), the incretin effect following oral glucose administration is reduced or absent (6,7), and the ability of intravenous GIP, but not GLP-1, to stimulate insulin secretion is severely blunted (7,8). This implies that a defective GIP signal transduction system and/or a reduced number of functional GIP receptors may contribute to the pathophysiology of type 2 diabetes. A greater understanding of the signal transduction systems activated by GIP should assist in determining whether reduced responsiveness involves changes at this level.
The receptor for GIP (9 -11) is a member of the class II G protein-coupled receptor superfamily, which includes receptors for glucagon, GLP-1, secretin, and vasoactive intestinal polypeptide (12). Stimulation of the GIP receptor has been shown to stimulate adenylyl cyclase and elevate intracellular cAMP levels in pancreatic islets (13), islet tumor cell lines (14), and various cell lines transfected with the GIP receptor (10,15,16). In addition, GIP has been shown to increase uptake of Ca 2ϩ into isolated islets (17) and increase intracellular Ca 2ϩ levels in HIT-T15 (18), RINm5F (9), and COS cells (10). We have shown that the GIP receptor probably couples to various Ca 2ϩ channels (10), but there is no evidence for GIP-stimulated IP 3 production (18). There is, however, evidence that GIP stimulates insulin secretion (19) and activation of mitogen-activated protein kinase (20) via a wortmannin-sensitive pathway, implying a role for phosphatidylinositol 3-kinase. It is therefore clear that GIP action on the pancreatic ␤-cell involves several interacting signal transduction pathways.
Heterotrimeric G proteins are activated by G protein-coupled receptors and undergo GDP/GTP exchange at the level of the G␣ subunit, leading to dissociation of the trimer into G␣ and G␤␥ subunits (33). The G␤␥ subunits have recently been shown to act on a number of effector targets including ion channels, enzymes, and kinases (33). Recent data suggested that inactivation of free G␤␥ completely abolished KCl, Ca 2ϩ , and GTP␥Sevoked insulin release from HIT-T15 cells (34), establishing a role for these subunits in insulin secretion. A role for G␤␥ has also been demonstrated in the coupling of PLA 2 and arachidonic acid production in rod outer segments (35) and to the activation of cardiac potassium channels (36).
These observations provided the rationale for determining whether arachidonic acid and PLA 2 are involved in the glucose potentiating effects of GIP in the ␤-cell, with a focus on G␤␥ subunits as a coupling mechanism. We show for the first time that GIP stimulates AA release from CHO-K1 cells and clonal ␤-cells (␤TC-3). Coupling of the GIP receptor to AA production in CHO-K1 cells was via G␤␥ subunits, whereas cAMP was shown to be the mediator in ␤TC-3 cells. The PLA 2 isoform activated by GIP in ␤TC-3 cells was Ca 2ϩ -independent and hypothesized to be the same as that activated by glucose when stimulating insulin secretion.

EXPERIMENTAL PROCEDURES
Cell Transfection and Tissue Culture-CHO-K1 cells cultured in Dulbecco's modified Eagle's medium/Ham's F-12 medium (Life Technologies, Inc.) and supplemented with 10% newborn calf serum (Cansera, Rexdale, Canada) were stably transfected with the wild type rat GIP receptor as described previously (10,15). The CHO-K1 cell line obtained by pooling clones was termed rGIP-15 and has previously been shown to express receptors at levels similar to high level expressing clones (15). In experiments targeted at investigating a role for G␤␥ signaling, rGIP-15 clones were transiently transfected with plasmid DNA encoding the C terminus of ␤-adrenergic receptor kinase (␤ARKct) (37) or the empty vector (pRK5). Briefly, 40 -60% confluent monolayers in 10-cm culture plates (Becton Dickenson, Lincoln Park, NJ) were transfected using Superfect TM (Qiagen, Valencia, CA) transfection reagent according to the manufacturers' protocol. Cells were harvested 18 -24 h posttransfection and passaged into 24-well plates for subsequent arachidonic acid release experiments. The empty plasmid pRK5 and the plasmid pRK-␤ARKct (495-689) were kindly provided by Dr. R. J. Lefkowitz (37). Passages 20 -30 of rGIP-15 cells were used in these experiments.
␤TC-3 cells were obtained from a frozen stock that was originally a gift from Dr. S. Efrat (Diabetes Center, Albert Einstein College of Medicine, New York) (38). Cells were cultured in low glucose (5.5 mM) Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 12.5% horse serum (Cansera) and 2.5% fetal bovine serum (Cansera). Passages 20 -30 were used in these experiments.
Iodination of GIP and Binding Analysis-Synthetic porcine GIP (5 g) was iodinated by the chloramine-T method, and the 125 I-GIP was further purified by reverse phase high performance liquid chromatography to a specific activity of 250 -300 Ci/g, (10). The aliquots were subsequently lyophilized and stored at Ϫ20°C until use. Competitive binding analyses were performed as described previously with minor modifications (10). Briefly, CHO-K1 cells plated 2 days prior in 24-well plates were washed twice with 4°C Krebs-Ringer (115 mM NaCl, 4.7 mM KCl, 1.2 mM KH 2 PO 4 , 10 mM NaHCO 3 , 1.28 mM CaCl 2 , 1.2 mM MgSO 4 ) containing 10 mM HEPES and 0.1% bovine serum albumin, pH 7.4 (KRBH), and incubated in triplicate for 14 -18 h at 4°C with 125 I-GIP (50 000 cpm/well) in the presence or the absence of unlabeled GIP (synthetic human GIP 1-42 ; Bachem, Torrence, CA). After two consecutive washes in ice-cold buffer, cells were solubilized with 0.1 M NaOH and transferred to test tubes for counting. Nonspecific binding was defined as that measured in the presence of an excess of human GIP (1 M), and specific binding was expressed as a percentage of maximum binding (%B/B o ).
cAMP and Insulin Determination-The cells were passaged into 24-well culture plates at 5 ϫ 10 4 cells/well for CHO-K1 clones and 5 ϫ 10 5 cells/well for ␤TC-3 cells. For cAMP studies, cells were washed twice with KRBH and then stimulated for 30 min with GIP in the presence of the phosphodiesterase inhibitor 3-isobutyl-1-methylxanthine at 0.5 mM concentration (RBI/Sigma). Following stimulation, reactions were stopped, and cells were lysed in 70% ice-cold ethanol, cellular debris was removed by centrifugation, and cAMP was subsequently quantified by radioimmunoassay (Biomedical Technologies Inc., Stoughton, MA). All insulin release experiments were performed over 60 min in KRBH in the absence of 3-isobutyl-1-methylxanthine, and insulin secreted into the medium was quantified by radioimmunoassay as previously reported (39).
Arachidonic Acid Release-Arachidonic acid release was determined by methods adapted from Shuttleworth and Thompson (40). Cells were harvested and passaged into 24-well culture plates at 4 ϫ 10 4 cells/well for CHO-K1 clones and 2 ϫ of experimental agents, the wells were washed twice with 0.5 ml of KRBH and allowed to equilibrate for 1 h. Ca 2ϩ -free experiments were conducted in KRBH containing equimolar Mg 2ϩ and supplemented with 10 mM EGTA. The agonists were dissolved in Krebs-Ringer buffer, added in triplicate (0.5 ml total volume/well), and incubated for the length of time shown in the figure legends. As a positive control, ATP was added at a final concentration of 5 M. When used, the inhibitor haloenol lactone suicide substrate (HELSS; Calbiochem, La Jolla, CA) was added for 30 min prior to washing and addition of agonists. After incubation, 0.4-ml aliquots were placed into scintillation vials followed by the addition of 10 ml of Econo 2 scintillation fluid (Fisher), and the radioactivity was determined by liquid scintillation spectrometry. AA released from cells was generally between 2-6% of total [ 3 H]AA incorporated into cells.
Data Analysis-The data are expressed as the means Ϯ S.E. with the number of individual experiments presented in the figure legend. All of the data were analyzed using the nonlinear regression analysis program PRISM (Graphpad, San Diego, CA), and the significance was tested using the Student's t test and analysis of variance (ANOVA) with the Tukey post-test (p Ͻ 0.05) as indicated in figure legends.

RESULTS
Initial studies were targeted at investigating GIP receptor signaling in an expression system, the rGIP-15 clone of CHO-K1 cells. Static incubations (45 min) revealed a concentration dependence to GIP-stimulated arachidonic acid production (Fig. 1a). In agreement with previous, non-incretin, studies on CHO-K1 cells (41,42), ATP (5 M) increased AA release from rGIP-15 cells by greater than 200% (p Ͻ 0.01, n ϭ 4). Parallel studies were performed in ␤TC-3 cells, a model of the pancreatic ␤-cell. These cells respond to arachidonic acid in a glucose-dependent manner (Fig. 2). In the presence of glucose, AA potentiated insulin secretion at concentrations as low as 10 M (Fig. 2a), whereas 20-fold greater concentrations were required before a response was observed under glucose-free conditions (Fig. 2c). The potentiation of insulin secretion elicited by 100 M AA is comparable with that elicited by 100 nM GIP under 11 mM glucose conditions (Fig. 2a versus Fig. 8). GIP was found to stimulate AA release in a concentration-dependent manner (Fig. 1, b and c). Interestingly, the EC 50 value for GIP-stimulated AA release (1.4 nM Ϯ 0.62 nM; n ϭ 3) was similar to that for insulin release in these cells (data not shown), in contrast to the 5-fold higher EC 50 value for cAMP production (39).
It is well established that the insulinotropic action of GIP is dependent on elevated glucose levels and that glucose induces activation of PLA 2 in pancreatic ␤-cells. However, we have recently shown that GIP receptor coupling to adenylyl cyclase in ␤TC-3 cells is independent of extracellular glucose concentrations (39). In the current study, increases in [ 3 H]AA efflux stimulated by GIP were also found to be independent of extra-  -3 cells (b). The medium was removed at indicated time points, and AA efflux was measured as described under "Experimental Procedures." Note that GIP-stimulated AA release was evident by 30 min; however, no effect of glucose was observed by this point. For a, n ϭ 4, and for b, n ϭ 3-4. *, p Ͻ 0.05. cellular glucose (Fig. 1c), indicating that GIP-induced and glucose-induced increases in AA release were mediated via separate pathways.
Analysis of the time dependence of AA release in rGIP-15 cells demonstrated maximal release at 10 min (Fig. 3a), which correlates well with that for GIP-stimulated cAMP production (maximal plateau reached at 10 -15 min in rGIP-15 and ␤TC-3 cells; n ϭ 3). In contrast, GIP-induced AA release was not detected before 30 min of incubation in the ␤TC-3 cells (Fig.  3b), and glucose-induced release was not observed until after 60 min of incubation (Fig. 3b). It was considered possible that these differences in onset of response may reflect alternative GIP receptor-effector coupling systems in the two cell types. Because G␤␥ has been previously implicated in the activation of phospholipase A 2 (35), an inhibitor peptide of G␤␥, ␤ARKct (␤-adrenergic receptor kinase C-terminal tail), was transiently expressed in rGIP-15 cells. To confirm that cells had been transfected, GIP receptor internalization was monitored because G␤␥ subunits have been shown to be required for G protein receptor kinase-mediated G protein-coupled receptor internalization (43). Expression of ␤ARKct was associated with an inhibition of receptor internalization in these cells (versus pRK5 vector control; n ϭ 3). Initial experiments were conducted to examine GIP receptor binding and cAMP production in this expression model. ␤ARKct expression was not found to have any significant effect on either receptor affinity for GIP or on activation of adenylyl cyclase (IC 50 values for binding: 3.95 nM Ϯ 0.91 (n ϭ 3) and 4.07 nM Ϯ 0.97 (n ϭ 3); EC 50 values for cAMP production: 0.73 nM Ϯ 0.12 (n ϭ 3) and 0.49 nM Ϯ 0.09 (n ϭ 3) for vector and ␤ARKct, respectively). GIP receptors were shown, for the first time, to be capable of functionally coupling to AA production through G␤␥ dimers, because the expression of ␤ARKct significantly suppressed the GIP-mediated response by almost 70% (Fig. 4, p Ͻ 0.05). Purinergic receptors were also found to be coupled to AA production via G␤␥ dimers, because ␤ARKct expression reduced ATP-stimulated AA production by greater than 40%.
To characterize further the pathway by which AA is produced in the ␤TC3-cell by GIP, the effect of ␤ARKct expression was investigated. To ensure that transfection had occurred, cells were typically cotransfected with green fluorescent protein as a marker of transfection efficiency. Inhibition of G␤␥ action had no effect on glucose-or GIP-stimulated AA release (␤ARKct expression versus pRK5 control; n ϭ 3) or insulin secretion in ␤TC-3 cells (Fig. 5). In addition, pertussis toxin (100 and 500 ng/ml) had no effect on AA release, indicating that toxin sensitive G␣-proteins (G␣ i , G␣ o , and G␣ q ) do not play a role in glucose-or GIP-stimulated AA release in ␤TC-3 cells (data not shown). This agrees with our previous studies showing that pertussis toxin at 500 ng/ml had no effect on cAMP levels in ␤TC-3 cells (39). However, both the diterpene forskolin and the incretin GLP-1, agents that specifically elevate intracellular cAMP levels, were able to stimulate AA release (Fig. 6), indicating that GIP may be acting on AA release via stimula-tion of adenylyl cyclase in the ␤TC3-cell. Interestingly, the specific protein kinase A inhibitor, H89 (5 M), showed no effect on GIP-or forskolin-stimulated AA release (Fig. 7). The possibility that AA-stimulated adenylyl cyclase activity can also be refuted because exogenous AA had either no effect or slightly inhibited basal cAMP production in ␤TC-3 cells (Fig.  2, b and d).
The reduction of extracellular Ca 2ϩ was found to have no effect on GIP-stimulated AA release, implying that a Ca 2ϩindependent mechanism was involved in the production of AA (Fig. 8a). As predicted, neither glucose nor GIP was able to stimulate insulin secretion from ␤TC3-cells under stringent Ca 2ϩ -free conditions (Fig. 8b). However, GIP was clearly still capable of elevating cAMP levels despite a reduction in basal cAMP production (Fig. 8c). The cAMP levels resulting from GIP stimulation under Ca 2ϩ -free conditions were, however, significantly suppressed compared with control conditions (p Ͻ 0.05). The ability of GIP to release AA under Ca 2ϩ -free conditions suggested that a Ca 2ϩ -independent PLA 2 was involved. An inhibitor specific for iPLA 2 , HELSS, has previously been shown to inhibit glucose-stimulated AA production and insulin secretion in several ␤-cell models (30,44,45). In the present study, HELSS was found to inhibit GIP-stimulated AA production as well as glucose-and GIP-stimulated insulin secretion (Fig. 9), supporting the aforementioned hypothesis that the enzyme coupled to GIP receptor signaling is a Ca 2ϩ -independent PLA 2 .

DISCUSSION
In human type 2 diabetes there is a decreased insulin response to GIP that is of unknown etiology. One possible underlying defect is in the normal signal transduction pathways by which GIP stimulates insulin secretion in ␤-cells. It has been established that GIP stimulates adenylyl cyclase (13), increases intracellular Ca 2ϩ (18), and activates mitogen-acti- FIG. 4. Effect of G protein ␤␥ inhibition on GIP-mediated arachidonic acid release in rGIP-15 cells. rGIP-15 cells expressing the GIP receptor were transiently transfected with 10 g pRK5 vector or ␤ARKct cDNA construct, and the medium was removed at 45 min. AA efflux was measured as described under "Experimental Procedures." The inset illustrates basal levels of AA release. *, p Ͻ 0.05.

FIG. 5. Effect of G protein ␤␥ inhibition on glucose and GIPpotentiated insulin secretion in ␤TC-3 cells. ␤TC-3 cells express-
ing the GIP receptor were transiently transfected with 10 g of pRK5 vector or ␤ARKct cDNA construct as described under "Experimental Procedures." Insulin secretion was assessed by radioimmunoassay and corrected for cell number by representation as the percentage of basal (n ϭ 3). *, p Ͻ 0.05 for basal versus all; #, p Ͻ 0.05 for 11 mM versus GIP; %, p Ͻ 0.05 for 10 nM GIP versus 100 nM GIP as tested by ANOVA. vated protein kinase (20), and the current study was undertaken to identify alternate mechanisms of regulating ␤-cell function. We have shown that GIP receptors in ␤TC-3 cells and transfected CHO-K1 cells are capable of coupling to transduction systems that release arachidonic acid from membrane lipids via activation of a Ca 2ϩ -independent phospholipase A 2 . Additionally, this signaling pathway was shown to involve G protein ␤␥ coupling in CHO-K1 cells, whereas a cyclic AMPmediated pathway is probably involved in ␤TC-3 cells.
Initial studies of GIP-stimulated AA release revealed a marked difference in the time dependence of AA release between CHO-K1 and ␤TC-3 cells. The much more rapid release evident in rGIP-15 cells is in agreement with previously observed AA production rates observed with rhodopsin and muscarinic receptors expressed in CHO-K1 cells (41,42) and other cell types (40,46,47). However, coupling of GIP to AA release was much slower in ␤TC-3 cells, suggesting a unique GIP receptor-AA coupling mechanism. There was also a difference between GIP-and glucose-induced AA release in ␤TC-3 cells, with GIP initiating release by 30 min, whereas glucose had no effect by this time (Fig. 3). This suggests that separate mechanisms couple glucose and the GIP receptor to AA production. Extensive studies have established that the glucose-induced AA production coupled to ␤-cell insulin secretion (24,48) involved activation of an ATP sensitive, Ca 2ϩ -independent PLA 2 (27,48). This enzyme has been identified in a number of insulinoma cell lines, including ␤TC-3 cells (44), and further studies were therefore performed to determine whether GIP-induced AA release also resulted from its activation.
The C-terminal fragment of the ␤-adrenergic receptor kinase protein (␤ARKct or G protein receptor kinase 2) was utilized to study the role of G␤␥ signaling. Jelsema and Axelrod (35) first suggested that activation PLA 2 can be performed by G␤␥ subunits. In the present study it was found that the GIP receptor can couple to PLA 2 via G protein ␤␥ subunits in CHO-K1 cells, whereas neither glucose nor GIP-stimulated arachidonic release or insulin secretion were dependent on G␤␥ subunit signaling in ␤TC-3 cells (Fig. 4). This is in contrast to their involvement in K ϩ -and bombesin-stimulated insulin secretion in HIT-T15 cells (34). Further studies are needed to determine whether G␤␥ subunits are involved in GIP receptor-effector coupling in other targets such as the stomach, fat, or the adrenal gland (49 -51).
Glucose-, GIP-, and ATP-stimulated AA release were all shown to be independent of extracellular Ca 2ϩ , indicating that they are likely acting on a similar iPLA 2 isoform. Recently cholecystokinin, another insulinotropic peptide, was also shown to activate islet PLA 2 independently of extracellular Ca 2ϩ (52). Despite a complete ablation of insulin release under Ca 2ϩ -free (extracellular) conditions, intracellular cAMP levels were still stimulated by GIP (Fig. 8c), implying that this may be the proximal messenger to AA release. A reduction in basal cAMP production is likely attributable to a decrease in basal Ca 2ϩ -activated adenylyl cyclase activity, therefore accounting for the reduction in GIP stimulated cAMP. From these observations it therefore seems likely that both GIP-stimulated cAMP and AA production are proximal signaling events independent of glucose and extracellular Ca 2ϩ but insufficient to elicit insulin exocytosis. However, these signaling intermediates may play more direct roles in the actions of GIP under euglycemic conditions, such as those in the adipocyte (50).
In islets, glucose stimulation can elevate endogenously generated AA from the micromolar range to cellular concentrations of 50 -200 M, as measured by mass spectrometry (48). In agreement with work published by Metz (53), exogenous AA over this range was able to stimulate insulin release from ␤TC-3 cells in the presence of glucose. However, in its absence, responsiveness to AA was reduced at least 10-fold. Interestingly, application of exogenous AA has been shown to elevate intracellular Ca 2ϩ concentrations in pancreatic islets (53), and there is considerable evidence suggesting a role for arachidonic acid itself or its metabolites in the regulation of capacitive and noncapacitive Ca 2ϩ influx in a number of cellular systems (54,55). Thus, it is tempting to speculate that fluxes in free endogenous AA, brought about by GIP, may play an integral role in regulating intracellular Ca 2ϩ concentrations and thereby influence insulin secretion.
Our studies indicate that the mediation of GIP-stimulated PLA 2 activity probably occurs via cAMP actions in the ␤-cell. Because the specific iPLA 2 inhibitor HELSS ablated insulin responses to glucose and thus the potentiating effect of GIP FIG. 8. Effect of Ca 2؉ -free extracellular media on GIP-mediated arachidonic acid release (a), insulin secretion (b), and cAMP production (c). Ca 2ϩ -free Krebs-Ringer buffer contained equimolar MgCl 2 to replace CaCl 2 and was supplemented with 10 mM EGTA. Arachidonic acid efflux (n ϭ 3-4), insulin (n ϭ 3), and cAMP levels (n ϭ 5) were determined as described under "Experimental Procedures." *, p Ͻ 0.05; **, p Ͻ 0.001. Gluc, glucose.
FIG. 9. Effect of Ca 2؉ -independent PLA 2 inhibition on GIPmediated arachidonic acid release (a) and insulin secretion (b) in ␤TC-3 cells. The cells were preincubated with the inhibitor, HELSS, for 30 min prior to stimulation and washed with KRBH before the addition of glucose and GIP. The inset in b represents basal insulin secretion levels under control and test conditions. *, p Ͻ 0.05. (Fig. 9) the converging actions on insulin secretion of these two secretogogues may occur distal to the formation of cAMP (by GIP) and arachidonic acid (by glucose and/or GIP). Arachidonic acid and/or its metabolites may therefore be mainly involved in the fine tuning of the insulin response. The actions of cAMP could be direct, via activation of small G proteins (e.g. Rap), or through a guanine nucleotide exchange factor; however, the involvement of protein kinase A is unlikely (Fig. 7). These results are in contrast to a recent study demonstrating an inhibitory affect of cAMP and incretins (GIP and GLP-1) on CCK-8-stimulated arachidonic acid production and insulin release in the rodent islet (56). However, implicit in studies conducted with isolated islets is the existence of paracrine and endocrine interactions between ␣-, ␦-, and PP cells that contribute to a functional response. This may account for the different responses observed in the clonal cell line used in the present study.
Finally, in the current study, production of AA was assessed by measuring the total radioactivity secreted from ␤TC-3 cells. Although it has been shown in studies on tumor ␤-cell lines that a surprisingly small percentage of released radioactivity consists of metabolites (48), a recent study suggested a role for lipooxygenase-12 metabolites in ␤-cell function (57). Further studies need to be conducted to discriminate between AA and its metabolites produced by GIP stimulation of the ␤-cell.