Intra-islet GLP-1, but not CCK, is necessary for β-cell function in mouse and human islets

Glucagon-like peptide 1 (GLP-1) and cholecystokinin (CCK) are gut-derived peptide hormones known to play important roles in the regulation of gastrointestinal motility and secretion, appetite, and food intake. We have previously demonstrated that both GLP-1 and CCK are produced in the endocrine pancreas of obese mice. Interestingly, while GLP-1 is well known to stimulate insulin secretion by the pancreatic β-cells, direct evidence of CCK promoting insulin release in human islets remains to be determined. Here, we tested whether islet-derived GLP-1 or CCK is necessary for the full stimulation of insulin secretion. We confirm that mouse pancreatic islets secrete GLP-1 and CCK, but only GLP-1 acts locally within the islet to promote insulin release ex vivo. GLP-1 is exclusively produced in approximately 50% of α-cells in lean mouse islets and 70% of α-cells in human islets, suggesting a paracrine α to β-cell signaling through the β-cell GLP-1 receptor. Additionally, we provide evidence that islet CCK expression is regulated by glucose, but its receptor signaling is not required during glucose-stimulated insulin secretion (GSIS). We also see no increase in GSIS in response to CCK peptides. Importantly, all these findings were confirmed in islets from non-diabetic human donors. In summary, our data suggest no direct role for CCK in stimulating insulin secretion and highlight the critical role of intra-islet GLP-1 signaling in the regulation of human β-cell function.


Results
Expression of pancreatic GLP-1 in mouse and human islets. Emerging evidence supports the production of GLP-1 in pancreatic islets 3,4,12,13 . Prohormone convertase 1/3 (PC1/3), the enzyme responsible for GLP-1 cleavage from a proglucagon precursor, has been detected in rodent glucagon-producing cells, especially under β-cell stress conditions 2,14,15 . However, the number of α-cells producing GLP-1 in non-stressed conditions has generally been assumed to be relatively few. Here, we evaluate whether fully processed active GLP-1 can be detected in islets isolated from lean mice and non-diabetic human donors and quantify its expression patterns using an antibody specific for the processed and biologically active GLP-1  amide form that does not cross-react with glucagon peptide 3 . The challenge for α-cell quantification in rodents includes the low number of α-cells in the islet (<10% in mouse islets). To overcome this problem, we dispersed islets into single-cells right before staining to maximize the number of α-cells counted in each experiment. As expected, mouse islets had fewer α-cells compared to humans (Fig. 1A). We detected active GLP-1 in both mouse and human islet cells (Fig. 1B). Quantification analysis shows that approximately 50% of glucagon-positive cells co-express GLP-1 in lean mice ( Fig. 1C-G). Surprisingly, these bi-hormonal cells were near 70% in humans (Fig. 1C,H-K). Essentially all of the GLP-1 expressing cells were α-cells. The body mass index (BMI) of the donors of the human islets used ranged from 28.5-31.8 kg/m 2 (islet preparations 5-8, Suppl. Table 1), consistent with our previous study in which islet GLP-1 secretion is increased in obesity 4 . Also, dispersed mouse islet cells had a similar expression pattern as intact islets (Suppl. Fig. 1), suggesting that acute islet dispersion did not affect either glucagon or GLP-1 expression. Together, our results suggest that GLP-1 is produced in a high percentage of α-cells, especially in human islets.
Paracrine GLP-1 signaling is necessary for GSIS. To test whether α-cell-derived GLP-1 plays an important paracrine role in β-cell function, we first measured bioactive GLP-1 secreted into the islet media as detected by active GLP-1 ELISA. Human islets secreted over 10-fold more GLP-1 than mouse islets on a per islet basis ( Fig. 2A,B), consistent with the higher number of GLP-1-positive cells observed in Fig. 1 as well as with a recent work comparing mouse vs. human pancreatic GLP-1 16 . We then tested whether paracrine GLP-1 signaling is necessary for normal glucose-stimulated insulin secretion by performing static GSIS studies in the presence of exendin-(9-39) (Ex9), a specific GLP-1 receptor (GLP-1R) antagonist 17 . We found that Ex-9 blunted GSIS in both mouse and human islets (Fig. 2C). Insulin content did not change significantly across all conditions and we found similar GSIS results when secretion was normalized to percentage of insulin content (data not shown). These results demonstrate that intra-islet GLP-1 signaling is necessary for GSIS in humans. Paracrine CCK signaling is not necessary for GSIS. Although it is not classified as an incretin hormone due to a general lack of effect on in vivo insulin secretion, CCK has been proposed to regulate insulin secretion in isolated rat islets 21 . We, therefore, hypothesized that, like pancreatic GLP-1, intra-islet CCK could also regulate β-cell insulin secretion. To test this, we performed in vitro and ex vivo studies in mouse and human islets. Here, we confirmed that CCK secretion is higher from ob/ob compared to lean mouse islets 5 and show that non-diabetic human islets secrete small amounts of active, sulfated CCK peptide (Fig. 4A). Importantly, we used a radioimmunoassay specific for sulfated CCK peptides with no cross-reactivity to gastrin peptide (Alpco, based on assay developed in 22 ). Since CCK has been proposed to stimulate insulin release in rodents 21 , we asked whether intra-islet CCK signaling is necessary for β-cell function. To answer this, we performed ex vivo GSIS in the presence of proglumide, a non-selective CCK receptor antagonist 23 . In contrast to Ex9 (Fig. 2C), we found that proglumide did not affect GSIS (Fig. 4B). This finding was consistent even in the islets from obese mice, which had both higher CCK and insulin release (Fig. 4A,B). Similarly, proglumide did not inhibit GSIS in human islets (Fig. 4B). Therefore, these results demonstrate that paracrine CCK signaling is not necessary for β-cell glucose stimulated insulin secretion. CCK does not regulate GSIS. The small amount of CCK released from the islet could be a reason for the lack of effect on β-cell function in a paracrine manner. However, in humans, the ability of CCK to stimulate insulin secretion is controversial [24][25][26] and has not been previously tested in isolated islets. Therefore, we tested if www.nature.com/scientificreports www.nature.com/scientificreports/ pharmacologic levels of CCK could directly potentiate insulin secretion ex vivo. We performed static GSIS studies in the presence of 100 nM 21 of sulfated (pGlu-Gln)-CCK-8, a stable CCK analog peptide 27 , at either low or high glucose. Contrary to our expectations, we found that exogenous CCK did not affect GSIS in human or mouse www.nature.com/scientificreports www.nature.com/scientificreports/ islets (Fig. 5). We then tried a high concentration of (pGlu-Gln)-CCK-8, 1 µM. In a recent publication, Khan, et al. found that only at this 1 µM concentration of (pGlu-Gln)-CCK-8 was there a modest increase in GSIS in isolated mouse islets 28 . However, we were unable to replicate these findings in mouse islets and even at this high concentration (pGlu-Gln)-CCK-8 did not augment GSIS (Fig. 5C). As CCKR signaling can stimulate increases in intracellular calcium (Ca 2+ ) in rat islets and other cell types 21,29,30 and this could be the trigger for insulin secretion in a β-cell, we measured cytosolic Ca 2+ in β-cells of isolated mouse islets before and after the addition of 1 µM CCK. We did not see a rapid or significant increase in intracellular Ca 2+ levels in response to CCK as would be expected to trigger insulin secretion (Fig. 5D). Importantly, all cells increased cytosolic Ca 2+ after 1 µM acetylcholine (ACh), assuring that β-cell Ca 2+ signaling was not impaired 31 . We repeated our GSIS assay in a 2 hour static incubation in INS-1 cells and again did not see any augmentation of GSIS with 100 nM (pGlu-Gln)-CCK-8 treatment (Suppl. Fig. 3). To ensure that this was not an effect of the modified (pGlu-Gln)-CCK-8 peptide, we repeated the assay with sulfated CCK-8, a native CCK peptide 27 , and still did not stimulate insulin secretion in β-cells whether normalized to insulin content or DNA content (Suppl. Fig. 3). Collectively, our results show no evidence that CCK directly promotes insulin secretion.

Discussion
The gut peptide hormone GLP-1 has been widely studied for its beneficial effects on β-cell insulin secretion and blood glucose control 6 . CCK has also been documented to stimulate insulin release in rodents 21,28 and humans 26 . Here, we show that both hormones are secreted from mouse and human pancreatic islets but only GLP-1 acts locally within the islet to promote insulin release. Additionally, we demonstrate that while CCK expression is regulated by glucose, its receptor signaling is dispensable for GSIS. Our data highlight the critical role of paracrine GLP-1 within the islet to maintain glucose homeostasis. We find that CCK is expressed and dynamically regulated in human islets, but its paracrine role in islet function remains unknown. GLP-1 produced by L cells in the gut is classically thought to act as an endocrine hormone on β-cells through the circulation. We and others have shown that islet GLP-1 is increased during conditions of islet-stress 2,4,14,15,32 , likely through increased pro-glucagon transcription and PC1/3 expression 9 . PC1/3 is highly expressed in β-cells because it is required for pro-insulin processing 33 . However, previous studies have shown that pancreatic α-cells contain the cellular machinery necessary to synthesize and secrete GLP-1, including PC1/3 2,3,13 (and online dataset from Benner et al. 34 ), suggesting that GLP-1 could also be produced in non-stressed islets. Based on these data, we 9 and others 12,15 have hypothesized that α-cell-derived GLP-1 acts on β-cell GLP-1R 9 . Here, GLP-1R antagonist markedly inhibited GSIS in islets from lean mice as well as from non-diabetic human donors, demonstrating that GLP-1 paracrine signaling pathways are already active in non-stressed β-cells (Fig. 2B). This suggests that there is a sufficient amount of GLP-1 present locally within the islet to activate its receptor on the neighboring β-cell. Indeed, we detected active GLP-1 being synthesized and released from both mouse and human islets, with an even greater amount in human islets than previously appreciated (Figs. 1 and 2A).
Previous studies have similarly suggested a paracrine signaling between α and β-cells where α-cell products directly regulate β-cell function 35 . A recent study showed that β-cell function depends on a local intra-islet glucagon signaling in mouse islets 36 . Furthermore, it has been shown that β-cell GLP-1R is necessary for glucose homeostasis and is dependent on Gcg peptides being produced in the islet 10 . Our study supports these findings and provides new evidence of α-cell-derived GLP-1 acting within the islet to promote β-cell function. Interestingly, glucagon has been reported as a relative low-potency agonist of GLP-1 receptor 37 , including in primary rodent β-cells 36,38 . Although our present data show that human islets produce and release much more GLP-1 in comparison to lean mouse, we cannot account for the possibility that other α-cell-derived peptides (i.e. glucagon) are also acting on β-cell GLP-1R 36,38 . Nevertheless, there is a limited understanding of this local signaling in vivo and more research is needed to confirm the relative role of α-to-β-cell crosstalk in regulating glucose homeostasis in humans.
In the present study, we also addressed whether pancreatic CCK might influence islet function in a paracrine manner. Since CCK has been shown to stimulate insulin secretion in rat islets 21 and humans 26 , we predicted that islet-derived CCK would affect β-cell insulin secretion. However, blocking CCK receptors did not decrease GSIS, suggesting that intra-islet CCK does not potentiate insulin release via its receptors. Moreover, there was no insulinotropic action of CCK analog peptide in isolated islets. This result was a bit surprising, however most of the studies demonstrating a direct effect of CCK on GSIS were performed in rat islets 21,39 . Several other in vivo studies in mice and humans failed to demonstrate enhanced GSIS with CCK administration 25,27,40 . While the exact experimental conditions (CCK concentration, incubation methods) differ in much of the literature, we directly tested conditions previously used to demonstrate that CCK peptide at 1 μM enhanced GSIS in mouse islets and could not replicate those findings (Fig. 5C) 28 . Interestingly, our live-cell imaging results suggest that CCK peptide does not activate robust increases in intracellular Ca 2+ in mouse β-cells as it does in pancreatic acinar cells 30 or vagal afferent neurons 29 and in rat islet 21 , and this lack of an effect on calcium influx may explain the inability to stimulate insulin secretion. In humans, both CCK receptor antagonists and CCK peptides did not affect insulin release ex vivo (Figs. 4B and 5B) or in vivo 41,42 , suggesting that CCK does not act as physiological incretin hormone. However, as CCK receptors are also expressed in other tissues, the peptide may be involved in the neural regulation of insulin secretion 43 , reduction of hepatic glucose production 11 and control of food intake 27 . CCK may also play a role in other regulated insulin secretion, such as amino acid-stimulated insulin secretion 44 or the enhancement of GSIS by GLP-1 45 . These beneficial effects, along with the impact of CCK on β-cell survival 4,5,46 , still make CCK a strong therapeutic candidate with the potential to treat obesity and type 2 diabetes 47 .
With increasing data indicate that classical gut hormones are produced locally in the islet 4,13,48,49 , many questions remain to be answered. For example, the molecular signals that turn on GLP-1/CCK expression in islet-cells www.nature.com/scientificreports www.nature.com/scientificreports/ remain to be fully elucidated. Also, a more detailed analysis of GLP-1/CCK-expressing cells in both mouse and human islets is warranted given there are several species differences. Both hormones act not only locally, but also systemically, increasing the complexity of understanding their overall impact on glucose homeostasis and beta-cell mass regulation. We have previously shown that overexpressing CCK in mouse β-cells does not change exocrine pancreas histology, nor systemic circulation of CCK peptide, suggesting that islet-derived hormone likely does not contribute significantly to systemic levels 46 . Similarly, recent studies showed that pancreatic GLP-1 does not necessarily change systemic GLP-1 circulating levels in mice 10,50,51 . These findings enhance our understanding on the biology of incretins and may open new possibilities for therapeutic innervations, including directed therapies to activate local pathways and avoid systemic side effects.
Overall, our results point to a physiological significance of paracrine GLP-1 but not CCK signaling in promoting glucose-stimulated insulin secretion in both mouse and human islets, suggesting that α to β-cell communication through GLP-1 receptor signaling is critical in the control of islet function and glucose homeostasis. Paracrine signaling of CCK is likely more important in β-cell survival pathways than in β-cell function and may be dynamically regulated by hyperglycemia.

Methods
Animals, islet isolation and culture. Animal care and experimental procedures were performed with approval from the Institutional Animal Care and Use Committee from the University of Wisconsin (protocol M005210) and William S. Middleton Memorial Veterans Affairs (protocol DD0001) to meet acceptable standards of humane animal care. All experiments were carried out in accordance with their guidelines and regulations. All animals used in this study were housed in facilities with a standard light-dark cycle and fed ad libitum. Pancreatic mouse islets were isolated from male 12-16 week old C57BL/6J and B6.Cg-Lep ob /J (ob/ob) (Jackson Laboratory, ME) as described previously 52 . Briefly, islets were isolated by collagenase digestion and Histopaque gradient (Sigma, #10771). Then, islets were handpicked and cultured at 37 °C and 5% CO 2 in RPMI 1640 media (Thermo Fisher Scientific, #11875093) containing 11 mM glucose and supplemented with 5 g/l BSA fraction V (Roche, #107351080001), 100 units/ml penicillin and 100 μg/ml streptomycin (1% P/S) (Thermo Fisher Scientific).
Human islets. Human islets were obtained through the Integrated Islet Distribution Program (https://iidp. coh.org/) (Suppl. Table 1). Upon arrival, islets were handpicked and then cultured in RPMI 1640 media (Thermo Fisher Scientific, #11879020) containing 8 mM glucose and supplemented as described above. Islets were cultured up to 7 days and the media was renewed every other day. mRNA levels. Gene expression analysis was performed as previously described 4,46 . RNA was isolated using either TRIzol or the QIAGEN RNeasy mini kit according to the manufacturer's instructions. cDNA was prepared using Applied Biosystems High Capacity cDNA synthesis kit, then analyzed by quantitative PCR (qPCR) using Power SYBR Master Mix (Life Technologies). All transcripts were normalized to β-actin. mRNA levels are shown as −ΔC T (CT[housing keeping] − CT[interest gene]). Fold changes were calculated relative to controls. Islet immunohistochemistry. Islets were dispersed at 37 °C using 0.25% Trypsin-EDTA (Thermo Fisher Scientific, #25200056), plated on Poly-L-lysine pre-coated glass coverslips and fixed with 10% formalin. Samples were permeabilized, blocked (Dako, #X0909) and then incubated with rabbit anti-glucagon (Santa Cruz, 1:200 -discontinued) and mouse anti-GLP-1 (7-36) amide (Abcam #ab26278, 1:200) antibodies. Abcam #ab26278 anti-GLP-1 (7-36) amide antibody recognizes the amidated C-terminus of the GLP-1(7-36) peptide and shows no cross-reactivity with non-amidated GLP-1, GLP-2, glucagon, or GIP 53 . Anti-rabbit Alexa Fluor 594 (1:400) and anti-mouse Alexa Fluor 488 (1:400) were used as secondary antibodies. Imaging was performed using an EVOS FL microscope. Manual scoring of images for co-localized glucagon, GLP-1, and DAPI (Dako) was performed using ImageJ (NIH) or Photoshop (Adobe) of, at least, 9 randomly chosen fields per treatment group for each replicate. DAPI staining was used as a nuclear marker while glucagon and GLP-1 were set as cytoplasmic stains.
Hormone secretion and assay. Islet GLP-1 and CCK secretion were previously described 4 . Briefly, islets were incubated at a density of 1 islet per 10 µl of media. For GLP-1 secretion studies, the DPP4 inhibitor (Millipore) was added to the media to prevent GLP-1 degradation 54 . A membrane permeable cAMP analog (8-CPT-cAMP, referred to as cAMP) was used to stimulate CCK secretion as previously described 4 . After incubation, media was collected, centrifuged at 1000 rpm for 5 min at 4 °C and supernatant stored at −80 °C until further analysis. GSIS was carried out as described previously 52 . Briefly, 45-60 islets were pre-incubated for 30 minutes in 0.5 mM glucose at 37 °C. Batches of 5 islets were then incubated for 1 hour at 37 °C in 24-well plates with different glucose concentrations and chemical compounds such as, exendin-(9-39) (Ex9) (Tocris Bioscience, #2081), proglumide (Tocris Bioscience, #1478), sulfated CCK-8 peptide (Cayman Chemical, #23371), and sulfated (pGlu-Gln)-CCK-8 (American Peptide). More details about glucose concentrations and compounds utilized can be found in each figure legend.
GLP-1 was measured using an Active GLP-1 ELISA (Millipore, #EGLP-35K). This ELISA does not detect any other forms of GLP-1, GLP-2, or glucagon (manufacturer's product information). Sulfated CCK levels were measured by radioimmunoassay with no cross-reactivity to the highly similar gastrin peptides (Alpco Diagnostics, now discontinued) as described 22 . Insulin was measured using an in-house insulin ELISA 55 , and QuantiFluor ® dsDNA System (Promega, #E2670) was used to measure islet DNA.
Live-cell imaging. Lean mouse islets were dispersed into small cell clusters using trypsin and gentle pipetting. Clusters were plated onto 35 mm culture dishes with cover glass bottom (Word Precision Instruments, #FD35) and cultured at 37 °C in the presence of 5% CO 2 with RPMI 1640 medium containing 8 mM glucose,