Multivalent activation of GLP-1 and sulfonylurea receptors modulates (cid:2) -cell second-messenger signaling and insulin secretion

Multivalent activation of GLP-1 and sulfonylurea receptors modulates (cid:2) -cell second-messenger signaling and insulin secretion. Am Physiol Cell Physiol 316: two pharmacophores that bind different cell surface receptors into a single molecule can enhance cell-targeting speciﬁcity to cells that express the complementary receptor pair. In this report, we developed and tested a synthetic multivalent ligand consisting of glucagon-like peptide-1 (GLP-1) linked to glibenclamide (Glb) (GLP-1/Glb) for signaling efﬁcacy in (cid:2) -cells. Expression of receptors for these ligands, as a combination, is relatively speciﬁc to the (cid:2) -cell in the pancreas. The multivalent GLP-1/Glb increased both intracellular cAMP and Ca 2 (cid:3) , although Ca 2 (cid:3) responses were signiﬁcantly depressed compared with the monomeric Glb. Moreover, GLP-1/Glb increased glucose-stimulated insulin secretion in a dose-dependent manner. However, unlike the combined monomers, GLP-1/Glb did not aug-ment insulin secretion at nonstimulatory glucose concentrations in INS 832/13 (cid:2) -cells or human islets of Langerhans. These data suggest that linking two binding elements, such as GLP-1 and Glb, into a single bivalent ligand can provide a unique functional agent targeted to (cid:2) -cells. ) and ATP-binding cassette subfamily C member 8 ( ABCC8 ) was relatively speciﬁc to the pancreas and (cid:2) -cells. RNA sequencing from the Human Protein Atlas (49; A ), Genotype-Tissue Expression (GTEx; B ), and Functional Annotation of Mammalian Genomes 5 (FANTOM5; 28; C ) data sets showed that the combinatorial expression of GLP1R and ABCC8 was relatively speciﬁc to pancreatic tissue. However, the GLP1R and ABCC8 expression combination was also present in the brain and heart. Single-cell RNA sequencing by Segerstolpe et al. (42) showed that combinatorial expression of GLP1R and ABCC8 was restricted to (cid:2) -cells ( D ) within the pancreas ( D – H ). RPKM, reads per kiolbase of transcript per million mapped reads; TPM, transcripts per million. The GTEx Project was supported by the Common Fund of the Ofﬁce of the Director of the National Institutes of Health and by the National Cancer Institute, National Human Genome Research Institute, National Heart, Lung, and Blood Institute, National Institute on Drug Abuse, National Institute of Mental Health, and National Institute of Neurological Disorders and Stroke. [Data used for tissue analysis ( A – C ) were accessed via the Human Protein Atlas (version 18; www.proteinatlas.org/ENSG00000112164-GLP1R/tissue and www.proteinatlas.org/ENSG00000006071-ABCC8/tissue), permission the Commons Attribution-ShareAlike International insulin concentrations were measured with the High Range Rat Insulin ELISA Kit (ALPCO) per the manufacturer’s instructions. Samples were run in quadruplicate and averaged. Insulin secretion rates are presented as a percentage of the dynamic range (%dynamic range) of secretion measured within each experiment; that is, basal insulin secretion in the presence of 1 G (min) and the maximum cAMP-potentiated GSIS induced by 15.5 mM glucose (cid:3) 10 (cid:7) M forskolin (max) were measured for each experiment, and the %dy-namic range for a given condition ( x ) was then calculated as a degree of for normalizing between individual experiments absolute as %dynamic range, these differences were reduced. Insulin secretion values were taken from at least eight replicate wells for each condition evaluated. the initial 6 min of the perifusion where islets were exposed to 2.8 G alone AUC of 6 min of 2.8 G (cid:3) ligand(s) period (11–17 AUC values for 16.7 G (cid:3) ligand(s) were calculated by subtracting the AUC of 12 min of the 16.7 G stimulation period of parallel experi- ments from the AUC of 12 min of the 16.7 G (cid:3) ligand(s) period (23–35 min). Statistical analysis. Data are presented as means (cid:9) SE. Statistical differences were determined using an unpaired two-sided Student’s t -test or a one-way ANOVA with a Tukey multiple-comparison correction in GraphPad Prism. A value of P (cid:11) 0.05 was considered statistically signiﬁcant for all statistical tests. There was no signiﬁcant difference in cAMP production when (cid:2) TC3 cells were exposed to GLP-1/Glb or monomeric GLP-1 across a range of concentrations. This suggests that the signaling induced by the GLP-1 moiety of the heterobivalent GLP-1/Glb ligand was preserved. Data represent means (cid:9) SE of well replicates (GLP-1/Glb: n (cid:12) 2–5 experiments, n (cid:12) 12–34 wells per concentration; monomeric GLP-1: n (cid:12) 2–5 experiments, n (cid:12) 10–30 wells per concentration). An unpaired two-sided t -test was used to compare cAMP production response at each concentration, and no signiﬁcant differences ( P (cid:11) 0.05) in (cid:2) TC3 cAMP production were observed between GLP-1 and GLP-1/Glb.

Most natural peptide hormones and biological signaling agents bind to a receptor on the cell surface to activate downstream signaling events. Many of these interactions involve a single recognition element within the ligand that binds a complementary region on its receptor in a monovalent interaction. However, some agents exhibit higher specificity for their receptors by interacting with multiple sites on the receptor's extracellular face, where one domain is required for activation and a second or multiple domains provide allosteric properties (21). GLP-1 exhibits multivalent binding to its receptor, wherein binding to a transmembrane domain is required for receptor activation while other sites on GLP-1 bind to the extracellular tail region of GLP-1R conferring allosteric regulation (24,25).
Higher-order multivalent interactions also have been demonstrated by coupling multiple copies of a ligand into a single molecule. In seminal studies, Sharma et al. developed melanotropin (MSH) conjugates containing multiple copies of a potent melanotropin analog ([Nle 4 ,D-Phe 7 ]␣-MSH; 43). These MSH analogs exhibited enhanced binding affinities, which were attributed to simultaneous interactions with multiple MSH receptors on the cell membrane (43). The effects of monovalent and multivalent homomeric MSH analogs on downstream signaling were also evaluated. The sensitivity for activation of cAMP formation was enhanced~100-fold for trivalent MSH compounds compared with the monovalent control suggesting that simultaneous binding to multiple receptors not only enhanced binding affinity but also activated downstream signaling efficiently (3).
On the basis of theoretical modeling, the increased affinity of multivalent ligands (MVLs) is likely due to proximity of the second binding domain to the membrane once the initial ligand of the complex binds to its receptor. The enhanced proximity greatly increases the probability for binding to a second receptor and, once bound, reduces the likelihood of dissociation thereby yielding high-affinity interactions (4). Thus, cell type specificity could be achieved by linking different elemental binding domains [receptor recognition elements (22)] into heteromultivalent ligand (htMVL) complexes that allow them to simultaneously bind different receptor types (3,19,50,53). Portoghese et al. evaluated a series of bivalent ligands composed of ␦and -opioid antagonists linked through a range of spacer lengths (36). Pharmacophores linked with a spacer of 21 atoms displayed the greatest enhancement in affinity (36). These studies concluded that the opioid receptors for the two unique ligands organized into functional heterodimers to obtain the enhanced binding (6,36). We subsequently proposed that if the expression of a "receptor cohort" is unique to a cell type of interest, a htMVL could be developed with recognition elements for a cell type-specific receptor combination, which would enhance target specificity (19,53). To test this possibility, a construct was synthesized with recognition elements for the melanocortin-1 receptor linked to a binding domain for the cholecystokinin B receptor. Again, a significant enhancement in binding affinity was observed, and this high-affinity binding was observed only to cells that expressed both complementary receptors (19,52). Specific targeting of this MVL to dual receptor-expressing tumors was then demonstrated in vivo, indicating the targeting potential of these agents (52). However, the aforementioned studies were performed on engineered cell lines that overexpressed the complementary receptor pair (19,52,53). Thus, we sought to characterize MVL binding and signaling in cells that were not engineered to overexpress the receptor pair of interest. To target pancreatic ␤-cells, we created a GLP-1 analog where GLP-1  was linked to the ␣ 2 -adrenergic receptor antagonist yohimbine. This GLP-1-yohambine htMVL was observed to bind at low concentration (1-5 nM) only to cells naturally expressing the complementary receptors and targeted islets of Langerhans in rodents (44). These findings demonstrated that htMVL can bind with high avidity and enhanced specificity to cells with endogenous expression of the complementary receptors.
Other unique dual ligands based on GLP-1 have been synthesized and tested for their efficacy in modulating ␤-cell function. Tschöp and colleagues synthesized estrogen coupled directly to a GLP-1 analog (11). In this case, GLP-1 was used to ferry the estrogen to any cell that expressed the GLP-1 receptor, but this ligand was unlikely to exhibit simultaneous binding to the complementary receptors; rather, the estrogen was expected to be released to the cytosol after cleavage from the construct within the lysosomal pathway. This approach demonstrates one of the potential uses of multivalent agents as cell-specific targeting/delivery agents. In other studies, Glb has been linked to glucose to improve the hydrophilicity of Glb and its usage as a targeting agent for imaging (41). This ligand exhibited high-affinity SUR binding with short-duration effects on blood glucose due to enhanced clearance (41). Recently, a bivalent cholecystokinin-GLP-1 analog was also produced and was found to activate insulin secretion from ␤-cells in vivo (18). Although intriguing, little is known regarding the downstream signaling that accompanies simultaneous activation of multiple pathways by a MVL.
To investigate the potential of a htMVL to uniquely modulate ␤-cell function, we synthesized a htMVL composed of the active fragment of GLP-1 {[Arg 36 ]GLP-1  } linked to Glb, a small organic agonist of SUR1 (GLP-1/Glb), and evaluated its effects on ␤-cell signaling and insulin secretion. The GLP-1 fragment was coupled to Glb using polyethylene glycol to attach both binding domains to a central string of prolineglycine repeats (14). GLP-1 and Glb were chosen because as monomers they are currently used as antidiabetic agents, and as a pair, their receptors exhibit relatively specific islet expression ( Fig. 1; 2 , 5, 13, 28, 30, 42, 45, 49). In our initial characterization, we reported the synthetic approach and examined the binding characteristics of this novel construct (14). A binding analysis with ␤TC3 cells indicated that the htMVL GLP-1/Glb exhibited two binding affinities with apparent K d of~10 and 60 nM. When GLP-1/Glb binding was evaluated in the presence of either of the unlabeled monomers, the high-affinity binding was lost, indicating that GLP-1/Glb binding in bivalent mode was enhanced~5-fold from the highest-affinity single binding element. Therefore, at concentrations below~20 nM, GLP-1/ Glb was predicted to exhibit specificity to cells that express both receptors, i.e., ␤-cells. Since the binding domains of these MVLs were directed against two different receptor types, GLP-1R and an ATP-binding cassette subfamily C transporter (ABCC8/SUR1), the ability of GLP-1/Glb to modulate downstream signaling pathways at the cellular level was of interest. In the present study, we focus on the ability of GLP-1/Glb to modulate intracellular Ca 2ϩ , cAMP, and insulin secretion relative to its monomeric components in ␤-cell lines and human islets of Langerhans.

MATERIALS AND METHODS
Ligand synthesis and purification. Details of ligand synthesis and purification have been presented previously in detail (14). Briefly, a carboxy-glibenclamide derivative was prepared in six synthetic steps from methyl-5-choro-salicylate (14). A truncated GLP-1 (residues 7-36) [GLP-1 ] was synthesized on a resin with solid-phase chemistry, modified at Lys 26 with a flexible oligoethylene glycol, and then conjugated to the carboxy-glibenclamide derivative. A thiol intermediate, GLP-1/Glb, was cleaved from the resin, purified by HPLC, and characterized by high-resolution mass spectrometry. Similarly, monomeric GLP-1  used in subsequent experiments was prepared by solid-phase chemistry on a rink resin, cleaved from the resin, purified by HPLC, and characterized by mass spectrometry.
Cytosolic Ca 2ϩ measurements. ␤TC3 cells were grown on 25-mm, no. 1 coverslips housed in six-well plates to~70% confluency. For cytosolic Ca 2ϩ concentration ([Ca 2ϩ ]i) measurements, coverslips with ␤TC3 cells were washed with HBSS at 37°C for 10 min and then loaded with 2.5 M of the acetoxymethyl ester (AM) form of fura-2 (Molecular Probes) with 0.0025% pluronic acid in HBSS-HEPESbuffered salt solution containing (in mM) 0.3 KH 2PO4, 138 NaCl, 0.2 NaHCO3, 0.3 Na2HPO4, 20 HEPES, 1.3 CaCl2, 0.4 MgSO4, and 5.5 glucose (unless otherwise noted) at pH 7.4 for 20 min at 37°C. Cells were rinsed in HBSS for 20 min at 37°C to allow for hydrolysis of the fura-2 AM. The coverslip with the dye-loaded cells was then placed in a chamber held at 37°C while mounted on the stage of an inverted Olympus IX-70 microscope equipped with a ϫ40 1.4-numerical aperture ultrafluar objective and a 150-W Xe lamp as the excitation source. Fura-2 was alternately excited at 340 and 380 nm using a filter wheel. The emitted light was filtered at 510 nm (10-nm band pass) and captured with a charge-coupled device camera (CH-350; Photometrics). Control images were acquired before the addition of the selected ligand, GLP-1/Glb or Glb (glybenzcyclamide; Alfa Aesar, Ward Hill, MA). Following ligand addition, images were acquired at regular intervals from 30 s to 10 min. Following the 10-min experimental period, 10 M ionomycin was added to the media, and images were acquired. On average, groups of 16 -20 cells were analyzed for each coverslip, and the average response from these cells was considered a single experiment. A minimum of five independent coverslips were analyzed for each experimental condition. For quantification, regions of interest were drawn within the cytosol of the cell using ImageJ (National Institutes of Health, Bethesda, MD), and the average pixel intensity was measured at 340 and 380 nm throughout the sequential series of images. For data analysis, fluorescence values were converted to a ratio of excitation at 340 and 380 nm and, using an in vitro calibration of fura-2, subsequently converted to [Ca 2ϩ ]i, as previously described (29,31). The control image for baseline [Ca 2ϩ ]i was subtracted from all subsequent images in a time series. The maximum GLP-1/Glb-and Glb-induced [Ca 2ϩ ]i responses were calculated, and data are presented as the maximal change in [Ca 2ϩ ]i over baseline during the experimental course. EC 50 and r 2 values were obtained using nonlinear regression analysis in GraphPad Prism (La Jolla, CA). cAMP assays. ␤TC3 cells were seeded in a 96-well plate at a density of 70,000 cells per well in RPMI with 10% FBS and 1% penicillin-streptomycin and cultured for 72 h. Cells were rinsed in serum-free RPMI and then cAMP assay medium [serum-free RPMI with 1 mM isobutyl-1-methylxanthine (IBMX)] for 5 min at 37°C. Cells were incubated with ligands in fresh cAMP assay medium for 15 min at 37°C. After treatment, media were discarded, cells were lysed, and the amount of cAMP was determined using a cAMP Chemiluminescent Immunoassay Kit (Invitrogen, Carlsbad, CA) per the manufacturer's instructions. EC50 and r 2 values were obtained using nonlinear regression analysis in GraphPad Prism. At least 10 wells per ligand concentration were analyzed, with up to 5 ligand concentrations evaluated per experiment.
Insulin secretion assays. Insulin secretion assays were performed using the INS 832/13 cell line, which has been shown to be an incretin-sensitive model with robust glucose-stimulated insulin secretion (GSIS; 17,38). INS 832/13 cells were seeded into a 24-well tissue culture plate at a density of 600,000 cells per well and cultured for 96 h. After 72 h, media were replaced with RPMI media containing 1 mM glucose (1 G). Insulin secretion for each well was measured in 1 ml of oxygenated HBSS with 1 G for 40 min at 37°C with gentle agitation. After this initial period, one-half of the media was removed, and an equal volume of oxygenated HBSS containing the desired glucose concentration Ϯ modulator was added. The incubation continued for an additional 40 min at 37°C; then media were collected, and insulin concentrations were measured with the High Range Rat Insulin ELISA Kit (ALPCO) per the manufacturer's instructions. Samples were run in quadruplicate and averaged. Insulin secretion rates are presented as a percentage of the dynamic range (%dynamic range) of secretion measured within each experiment; that is, basal insulin secretion in the presence of 1 G (min) and the maximum cAMP-potentiated GSIS induced by 15.5 mM glucose ϩ 10 M forskolin (max) were measured for each experiment, and the %dynamic range for a given condition (x) was then calculated as This protocol provided a high degree of precision for normalizing between individual experiments wherein absolute rates might vary greatly; when evaluated as %dynamic range, these differences were reduced. Insulin secretion values were taken from at least eight replicate wells for each condition evaluated.
The chambers containing the islets were maintained at 37°C. Islet response to 16.7 mM glucose (16.7 G) alone was measured in parallel with experimental perifusions. The perifusate was collected in an automatic fraction collector designed for a 96-well plate. Samples (100 l) were collected every minute and were stored at Ϫ80°C. Insulin secreted into each sample was measured using a human ELISA kit (no. 10-1113; Mercodia, Uppsala, Sweden) according to the manufacturer's instructions. If necessary, samples were diluted 1:10 in KRB to provide a concentration that fell within the range of the standard curve; otherwise, undiluted samples were used. Samples were run in duplicates. Insulin was normalized to the total DNA from the islets in each chamber. DNA was quantified using the Quant-iT PicoGreen dsDNA kit (Life Technologies, Carlsbad, CA) according to the manufacturer's instructions. Islets were diluted in 1.0 ml AT buffer (1 M ammonium hydroxide and 0.2% Triton X-100) and then sonicated for 15 s. Samples were run in quintuplicates. Experimental perifusion values are presented as a percentage of the maximum GSIS measured in parallel experiments. Area under the curve (AUC) values for 2.8 G ϩ ligand(s) were calculated by subtracting the AUC of the initial 6 min of the perifusion where islets were exposed to 2.8 G alone from the AUC of 6 min of the 2.8 G ϩ ligand(s) period (11-17 min). AUC values for 16.7 G ϩ ligand(s) were calculated by subtracting the AUC of 12 min of the 16.7 G stimulation period of parallel experiments from the AUC of 12 min of the 16.7 G ϩ ligand(s) period (23-35 min).
Statistical analysis. Data are presented as means Ϯ SE. Statistical differences were determined using an unpaired two-sided Student's t-test or a one-way ANOVA with a Tukey multiple-comparison correction in GraphPad Prism. A value of P Ͻ 0.05 was considered statistically significant for all statistical tests.

[Ca 2ϩ ] i changes were diminished by the multivalent GLP-1/Glb ligand.
Glb is known to elevate [Ca 2ϩ ] i in ␤-cells via its inhibition of ATP-sensitive K ϩ (K ATP ) channel activity and subsequent membrane depolarization. To measure the effect of GLP-1/Glb on [Ca 2ϩ ] i , we developed dose-response profiles of the maximal [Ca 2ϩ ] i change elicited by GLP-1/Glb and monomeric Glb (Fig. 2). Both GLP-1/Glb and Glb elicited concentration-dependent increases in [Ca 2ϩ ] i . However, the responses elicited by GLP-1/Glb were reduced over the entire concentration range compared with monomeric Glb, suggesting that the Glb moiety of the htMVL was not as efficacious as monomeric Glb (Fig. 2).
cAMP signaling of GLP-1/Glb was equivalent to that of GLP-1. GLP-1 is known to potentiate ␤-cell insulin secretion via elevation of cAMP, whereas Glb does not influence cellular cAMP levels, which we confirmed for the ␤TC3 cell line (Fig.  3). To investigate the signaling properties of the GLP-1 moiety of the GLP-1/Glb ligand, we compared cAMP production in ␤TC3 cells in response to several concentrations of GLP-1/Glb or monomeric GLP-1 (Fig. 4). Significant increases in cAMP production were observed at 1 nM for both the GLP-1/Glb ligand and GLP-1 monomer. However, no significant differences were observed between GLP-1/Glb ligand and monomeric GLP-1 over a wide concentration range (GLP-1/Glb, EC 50  In the presence of stimulatory concentrations of glucose (15.5 G), GLP-1/Glb increased GSIS in a dose-dependent manner (Fig. 6). At 5 nM, the enhancement of GSIS by GLP-1     not significantly different (Fig. 6). Five nanomolar Glb alone did not potentiate GSIS; however, when added in combination as monomers at 5 nM, GLP-1 and Glb increased GSIS significantly and to levels greater than those for the heterobivalent GLP-1/Glb (Fig. 7). GLP-1/Glb modulation of human islet insulin secretion. We were particularly interested in investigating whether GLP-1/ Glb could modulate insulin secretion by human islets given that the heterobivalent ligand 1) did not induce insulin secretion at nonstimulatory glucose concentrations and 2) enhanced GSIS, albeit at levels lower than those for the combination of monomeric GLP-1 and Glb (Figs. 5 and 6). To this end, we performed a focused study evaluating the relative effects of the GLP-1/Glb ligand and the combination of monomeric GLP-1 and Glb ligands on isolated human islet insulin secretion (Fig.  8). At 2.8 G, 1 nM GLP-1 and Glb significantly increased insulin secretion relative to GLP-1/Glb (Fig. 8, A and C). At 2.8 G, both the combination of 5 nM GLP-1 and Glb and GLP-1/Glb induced an increase in insulin secretion (Fig. 8, A  and C). At 16.7 G, both the combination of GLP-1 and Glb and GLP-1/Glb increased insulin secretion at 1 and 5 nM concentrations. However, GSIS was enhanced~1.5-2-fold in the presence of 5 nM ligand(s) concentration compared with 1 nM ligand(s) concentration with an apparent elevation of secondphase insulin secretion (Fig. 8B).

DISCUSSION AND CONCLUSIONS
Here we present a functional characterization of a ␤-celltargeted htMVL composed of linked GLP-1  and Glb binding domains (14). Our findings demonstrate that the GLP-1/Glb htMVL differentially activated ␤-cell signal transduction and insulin secretion relative to monomeric GLP-1 and/or Glb.
An important consideration of GLP-1/Glb binding was that the individual binding moieties within the MVL had lower affinities for their respective receptors (GLP-1R and SUR1) compared with unconjugated monomers. We previously estimated that the binding constants for the individual recognition elements within the bivalent ligand were~80 nM for GLP-1  and~40 nM for Glb, and high-affinity binding of GLP-1/Glb (K d~5 nM) was observed only when both complementary receptors (i.e., GLP-1R and SUR1) were available (14). Thus, activity of GLP-1/Glb below~10 nM was presumed to be mediated by bivalent interactions, whereas activity at higher concentrations may have included monovalent ligand-receptor interactions of individual binding moieties within the GLP-1/Glb (particularly Glb-SUR1) htMVL (14).
Previous data have shown that activation of GLP-1R signaling combined with elevated glucose causes a more pronounced inhibition of K ATP channels [SUR1 and inward rectifier K ϩ channel 6. all concentrations of GLP-1/Glb tested compared with monomeric Glb (Fig. 2). These data suggest that the Glb moiety of the htMVL was not as effective as monomeric Glb. However, GLP-1/Glb was an effective and potent activator of cAMP production with an EC 50 similar to monomeric GLP-1 (Fig. 4) suggesting that GLP-1R second-messenger activation by the GLP-1/Glb was not impaired but inhibitory sensitization of the K ATP complex mediated by actors downstream of cAMP was. Previous data have also shown that interaction of sulfonylureas with exchange protein directly activated by cAMP 2 isoform A (Epac2A)/Rap1 signaling is crucial for combinatorial incretinsulfonylurea augmentation of insulin secretion (46,47). However, the ability of sulfonylureas to directly activate Epac2A is controversial (16). Given the complicated nature of incretin and sulfonylurea signaling, we predict that the observed reduction in [Ca 2ϩ ] i elicited by GLP-1/Glb was caused by multiple factors. First, the Glb moiety of GLP-1/Glb, when bound bivalently, did not inhibit the SUR1 as effectively as monomeric Glb because of its structural relationship with the polyethylene glycol linker and an engaged GLP-1-GLP-1R complex. At concentrations Ͻ100 nM, GLP-1/Glb was largely ineffective at changing [Ca 2ϩ ] i . However, as GLP-1/Glb concentration was raised above 100 nM, larger changes in [Ca 2ϩ ] i were observed (Fig.  2). Thus, at higher concentrations, we predict that the Glb moiety of the bivalent ligand bound to, and partially activated, additional SUR1 in a monovalent and less potent mode. Second, bivalent binding of GLP-1/Glb to GLP-1R and SUR1 did not appear to cause additional K ATP channel inactivation or increased [Ca 2ϩ ] i as observed by others (27,46). Our data suggest that this interference did not occur through inhibition of cAMP (Fig. 4) but most likely through interference between PKA and SUR1 and/or disruption of Glb activation of the Epac2A/Rap1 axis.
Insulin secretion in the presence of GLP-1/Glb was also altered relative to monomeric GLP-1 and/or Glb. At nonstimulatory glucose concentrations, INS 832/13 cells in the presence of the monomeric combination of 5 nM GLP-1 and Glb significantly elevated insulin secretion, whereas neither 5 nM monomeric Glb nor 5 nM monomeric GLP-1 alone activated insulin secretion, indicating a sensitization of SUR1 inhibition and/or activation of Epac2A/Rap1 signaling, as previously reported (27,46). Conversely, 5 nM bivalent ligand, at nonstimulatory glucose concentrations, had no significant effect on "basal" insulin secretion in INS 832/13 cells. In human islets, 1 nM bivalent GLP-1/Glb did not stimulate basal insulin secretion to the same degree as the monomeric combination of GLP-1 and Glb did, demonstrating that the bivalent ligand was not acting as a sum of its constituent monomers at low glucose concentrations, even though binding in bivalent mode was significant (14). On the other hand, 5 nM GLP-1/Glb did stimulate basal insulin secretion in human islets (to levels 75% of that observed for GLP-1 and Glb in combination) suggesting that increasing concentrations of GLP-1/Glb may be able to restore some level of GLP-1R-mediated inhibitory sensitization of the K ATP and/or Epac2A activation. Moreover, GLP-1/Glb and the monomeric combination of GLP-1 and Glb, at both 1 and 5 nM, enhanced GSIS to similar levels. Interestingly, at 5 nM ligand concentration, GLP-1 and Glb combined and GLP-1/Glb increased GSIS~2-fold relative to that observed at 1 nM ligand concentration because of apparent prolonged elevation of second-phase insulin secretion not observed at 1 nM. Combined, these data suggest that the htMVL GLP-1/Glb interrupted inhibitory sensitization of K ATP and/or Epac2A signaling at nonstimulatory glucose concentrations but augmentation of insulin secretion at elevated glucose concentrations was preserved compared with its constituent monomers.
Since the high-affinity binding of GLP-1/Glb was dependent on the availability of both receptors, our findings indicate that this bivalent ligand may, at concentrations below 20 nM, serve as an incretin targeted to ␤-cells. Importantly, GLP-1/Glb not only retained the capacity to promote GSIS but also had little effect on insulin secretion in the absence of stimulatory glucose at concentrations below 5 nM, contrary to its constituent monomers. Therefore, in addition to the potential for limiting off-target effects of GLP-1 agonists due to enhanced ␤-cell specificity, GLP-1/Glb may be useful in preventing hypoglycemic incursions during fasting periods while retaining considerable GLP-1 potency and potentiation of ␤-cell GSIS during nonfasted states. Further investigation of downstream signaling and optimization of heterobivalent ligands as a ␤-cell-specific therapeutic agent appear warranted. Moreover, the rules learned here guide further development for tuning both specificity and signaling properties of multivalent agents.

ACKNOWLEDGMENTS
We thank Renata Patek and Jane Zhang for the synthesis of the heterobivalent GLP-1/Glb and Dr. Christopher Newgard (Duke University, Durham, NC) for access to the INS 832/3 cell line. We also acknowledge the Integrated Islet Distribution Program at the City of Hope (Duarte, CA; https://iidp.coh.org/) for providing the human pancreatic islets for perifusion studies.