Heterogeneity in the coordination of delta cells with beta cells is driven by both paracrine signals and low-density Cx36 gap junctions

Significance

The regulated secretion of insulin and glucagon is exceptionally important for the maintenance of normal blood glucose, including the prevention of hypoglycemia. Local release of the inhibitory hormone somatostatin provides essential feedback to ensure that insulin and glucagon are released when appropriate. This feedback is significant since excess somatostatin release in T1D prevents the release of glucagon to counter deadly hypoglycemia. Conversely, excess insulin, delivered or secreted, can quickly induce life-threatening hypoglycemia. The mechanism for this crosstalk is unresolved. Here, we established that a combination of paracrine signaling and low-density Cx36 gap junctions between beta and delta cells are the underlying mechanisms, providing therapeutic opportunities to maintain or restore glycemic control in diabetes and congenital hyperinsulinemia (CHI).

Abstract

Insulin potently decreases blood glucose; thus, tight control is required to prevent excessive insulin release and hypoglycemia. Central to this inhibition is somatostatin released from delta cells that are clustered with beta cells in pancreatic islets. This communication is of interest because the loss of functional beta cells in diabetes leads to uncontrolled delta cell activity that disrupts islet paracrine crosstalk. While it is established that insulin and somatostatin secretion are coordinated, the specific mechanism is unsettled. We have previously demonstrated that beta cells release the hormone Urocortin 3 to stimulate delta cells at high glucose, demonstrating a paracrine negative feedback loop. Others have proposed direct coordination via gap junctions. To resolve this conundrum, we used the genetically encoded fluorescent Ca²⁺ reporter GCaMP6s to simultaneously record the activity of hundreds of beta and delta cells in low (2.8 mM) (LG) and high (16.8 mM) glucose (HG). Surprisingly, while many delta cells exhibit Ca²⁺ oscillations in HG that are coordinated with beta cells, the activation of these delta cells precedes beta cells and is more variable than beta cell responses. The selective delta cell knockout of connexin 36 confirmed the involvement of gap junctions. However, blockade of vesicle release with the Rho-GTPase inhibitor ML-141 completely removed coupling between beta and most delta cells in HG. Our data reveal considerable functional heterogeneity among delta cells, where most delta cells are entrained by oscillatory Ca²⁺ behaviors of beta cells that are mediated by a combination of paracrine signaling and low-density gap junction coupling.

Pancreatic islets are small clusters of hormone-secreting cells responsible for nutrient regulation in our blood, and are primarily composed of insulin-secreting beta cells, glucagon-secreting alpha cells, and somatostatin-secreting delta cells (1). Of particular importance to healthy islet function is the rapid and precise suppression of insulin secretion when blood glucose levels have returned to normal, as insulin’s actions on peripheral tissues lowers blood glucose (2). Insulin’s potent effect on blood glucose is exemplified in normal physiology by the early rise in both insulin and glucagon during the cephalic phase. Insulin prepares the body to receive and digest food, while glucagon ensures that the mere sight, smell, or taste of food does not induce hypoglycemia. In contrast, during exercise the release of catecholamines potently suppresses insulin secretion while promoting glucagon secretion to keep glucose levels stable. In the context of disorders of unregulated insulin secretion such as congenital hyperinsulinemia (CHI), excess insulin secretion leads to severe hypoglycemia if untreated (3). Additionally, inappropriately dosed exogenous insulin can result in iatrogenic hyperinsulinemic hypoglycemia, which can be imminently fatal (4). For these reasons, insulin is one of the 5 most common drugs leading to hospitalization in the United States. (5). Precisely because excess insulin is such a potent and dangerous hormone, our pancreas employs several complementary mechanisms to ensure appropriate restraint of insulin release at all times. This includes beta cell autonomous mechanisms like the expression of glucokinase, which acts as a glucose sensor preventing beta cell stimulation at low glucose concentrations, and gap junctions, which in addition to coordinating beta cells within the islet also act to suppress them at low glucose.

Additionally, paracrine inhibition via somatostatin released by delta cells acts to modulate insulin secretion by transiently suppressing beta cell secretion (6–9). Beta and delta cell secretion in response to high glucose is pulsatile and coordinated; somatostatin secretion trails insulin release by 30 s (10, 11), a time delay consistent with a mechanism of paracrine signaling. In response to elevated blood glucose, beta cells activate delta cells, in part via the secretion of paracrine factors such as Urocortin 3 (Ucn3) (12). The subsequent secretion of somatostatin attenuates beta cell insulin secretion and closes a negative feedback loop, ensuring a controlled return to euglycemia (13). The onset of Ucn3 expression and subsequent stimulation of somatostatin-mediated feedback inhibition by delta cells contributes to the glucose set point in mice (12, 14, 15). Conversely, beta cells downregulate the expression of Ucn3 early on in prediabetes—lessening the feedback inhibition of insulin secretion, presumably to adapt to increasing peripheral insulin resistance. In type 1 diabetes, excessive delta cell inhibition of alpha cells prevents counterregulatory glucagon secretion (16–18), which was recently linked to the loss of beta cells (19). Gap junctions connecting beta cells ensure coordinated depolarization and synchronous pulsatile insulin secretion. A less appreciated but important consequence of the gap-junction connections between beta cells is that they help prevent individual beta cells within intact islets from activating under low glucose (20–24). It has been reported that delta cells are coupled to beta cells through gap junctions, as optogenetic-stimulation of beta cells activates neighboring delta cells within an average delay of 30 ms, a time delay consistent with gap junction coupling (25). These gap junctions had similar conductance as the gap junctions connecting neighboring beta cells and were suggested to be Cx36, although the presence and identity of Cx36 gap junctions was not experimentally validated in a loss of function approach. However, delta cells are active at glucose concentrations as low as 3 mM (26, 27), below the threshold for beta cell insulin secretion. These observations indicate that gap junction connections between beta and delta cells are on their own unlikely to explain key elements of the carefully coordinated beta–delta behaviors that occur within intact pancreatic islets. Importantly, gap junction-mediated and paracrine coordinated activity of beta and delta cells upon glucose stimulation are not mutually exclusive mechanisms to explain the coordinated secretion of insulin and somatostatin release from intact islets.

To determine the underlying mechanisms of intercellular communication between beta and delta cells in response to elevated glucose, we measured Ca²⁺ dynamics of hundreds of delta cells and thousands of beta cells in intact islets. We combined these observations with a series of genetic and pharmacological approaches designed to separately perturb paracrine and gap junction-mediated communication. Through these experiments, we observe highly heterogeneous delta cell behaviors that can be used to categorize delta cells based on their activity in low and high glucose and their level of coordination with neighboring beta cells. We demonstrate the presence of a low density of Cx36 gap junctions on a subset of beta/delta connected plasma membranes, and validate through delta cell-selective deletion of Cx36 that these gap junctions are required for beta cells to entrain neighboring delta cells. In parallel, we observe that inhibition of paracrine secretion via application of the Cdc42 inhibitor ML-141 to islets interrupts oscillatory Ca²⁺ responses specifically in delta cells, with entrainment resuming upon ML-141 washout. Based on these observations, we conclude that delta cells demonstrate profound functional heterogeneity, and that their behavior is far more complex than to simply be explained by gap junction connections to neighboring beta cells alone. Instead, our results highlight that delta cells are capable of substantial beta cell-independent but depolarization-dependent Ca²⁺ behavior, particularly during low glucose. Upon high glucose stimulation, many delta cells become entrained by beta cells and coordinate their Ca²⁺ activity with neighboring beta cells via a combination of gap junction coupling and paracrine signaling, while other delta cells behave independently from beta cells. Our results establish how a combination of paracrine and electrical signals that result from beta cell activity form the mechanistic basis for the synchronization of delta cell activity with neighboring beta cells under high glucose stimulation.

Results

Beta and Delta Cell Ca²⁺ Responses Are Mediated by K_ATP Channels and L-Type Ca²⁺ Channels.

Delta cells, like beta cells, rely on K_ATP channels and L-type voltage gated Ca²⁺ channels for secretion (12, 28). To confirm and extend their role in mediating Ca²⁺ spikes in delta cells, we generated Sst-cre x lsl-GCaMP6s mice that selectively express the green fluorescent Ca²⁺ reporter GCaMP6s in delta cells. We then documented live Ca²⁺ response using a customized microfluidics-based perfusion system. At 5.5 mM glucose, below the threshold for beta cell activation, many delta cells demonstrated a robust pattern of random-like, short Ca²⁺ responses that are inhibited by the K_ATP channel opener diazoxide and the L-type Ca²⁺ channel blocker isradipine (SI Appendix, Fig. S1 A–D and Movie S1). As isradipine does not wash out quickly, delta cell Ca²⁺ activity did not immediately return. Of note, while the average delta cell response is significantly attenuated by diazoxide and isradipine, some delta cells continue to exhibit Ca²⁺ spikes. This likely reflects a depolarization-independent mechanism of intracellular Ca²⁺ mobilization (29). Since delta cells, like beta cells, are further activated by elevated glucose, we repeated this experiment in mIns1-H2Bb-mCherry x Sst-cre x Ucn3-cre x lsl-GCaMP6s mice, which express GCaMP6s in both beta and delta cells and mCherry exclusively in beta cell nuclei (30). In response to 16.8 mM glucose, beta and delta cell Ca²⁺ increased robustly and assumed a synchronous oscillatory pattern, in line with the reported synchronized release of insulin and somatostatin (10). The addition of diazoxide and isradipine each elicited a sharp decline in Ca²⁺ activity for both beta and delta cells (SI Appendix, Fig. S1 E–H and Movie S2).

Delta Cell Ca²⁺ Oscillations in High Glucose Are Coordinated With Beta Cells but Not Identical.

To further investigate beta and delta cell coordination within intact islets, we recorded Ca²⁺ response of islets isolated from mIns1-H2B-mCherry x Sst-cre x Ucn3-cre x lsl-GCaMP6s mice to 2.8 mM glucose (low glucose, LG), followed by 16.8 mM glucose (high glucose, HG). Cell identity was validated by the combination of beta cell nuclear mCherry expression, response to the delta cell specific activator ghrelin (26, 31), and a whole mount immunofluorescence post hoc stain for insulin and somatostatin of each islet (Fig. 1A and Movie S3). As before, in LG delta cells exhibited spontaneous, uncoordinated Ca²⁺ behaviors while beta cell Ca²⁺ activity was nonexistent. In HG, beta cells exhibited synchronized, pulsatile Ca²⁺ activity as expected (32). Delta cell Ca²⁺ activity under HG exhibited a much more complex pattern than the smooth oscillations of beta cells, with prominent fast oscillations that appear uncoordinated with beta cells, superimposed onto slow Ca²⁺ oscillations that are synchronous with beta cells. To better resolve the delta and beta cell Ca²⁺ behaviors, we wrote a custom Python script to separate the slow and fast components of beta and delta cell Ca²⁺ activity. Beta cell Ca²⁺ activity was largely composed of a slow component (period between 1 to 20 min). Any residual fast Ca²⁺ elements (period less than 1 min) predominantly overlapped with peaks in the slow component (Fig. 1B). In contrast, delta cell Ca²⁺ behavior under HG was composed of slow oscillations coordinated with beta cells and fast oscillations that largely occurred during nadirs in the slow component (Fig. 1C).

Fig. 1.

Delta cells exhibit both beta cell coordinated and uncoordinated Ca²⁺ oscillations. (A) Intensity plot of an islet expressing GCaMP6s in beta cells (green), and delta cells (red) in response to LG, HG, 100 nM ghrelin, and 30 mM KCl as a cell viability indicator with line graphs of several representative cells superimposed. Image of the islet during live cell imaging from Movie S3 (Top Right) and accompanying immunofluorescent stain (Bottom Right). (B and C) Line graph of a single beta cell (green, Left panel) and delta cell (red, Right panel) deconstructed into a baseline, slow component, fast component, and at the Bottom an overlay of the slow (orange) and fast (blue) components.

Delta Cell Ca²⁺ Flux Precedes Beta Cell Ca²⁺ Flux in Response to Glucose and KCl.

To our initial surprise, a careful inspection of beta and delta cell Ca²⁺ activity within the same islets revealed that the slow component of delta cell Ca²⁺ oscillations consistently preceded beta cell Ca²⁺ oscillations under both HG and direct depolarization with 30 mM KCl. To quantify these differences, we extracted four parameters for each Ca²⁺ oscillation or spike: the “start of pulse,” “start of plateau,” “end of plateau,” and the “end of pulse”. With these key parameters established, we quantified that the maximum slope of an average delta cell during the rise (start of pulse to start of plateau) and fall (end of plateau to end of pulse) of each Ca²⁺ wave was approximately 1.5x faster than beta cell Ca²⁺ responses under identical HG stimulation (Fig. 2 A–C). In response to KCl, delta and beta cell Ca²⁺ rise times were not statistically different, but delta cell fall times were about 1.3x faster (Fig. 2 D–F). We then conducted pairwise comparisons for each of the four parameters listed above between beta cells (beta–beta), beta and delta cells (beta–delta), and delta cells (delta–delta) in the same islets. In response to HG, delta cell Ca²⁺ responses for each parameter consistently preceded beta cell Ca²⁺ responses (start of pulse: 16.7 s; start of plateau: 48.5 s; end of plateau: 38.4 s; end of pulse: 107.8 s) (Fig. 2C and SI Appendix, Fig. S2 A–D). Delta cells also preceded beta cells under 30 mM KCl (start of pulse: 29.7 s; start of plateau: 31.6 s; end of plateau: 19.9 s; end of pulse: 123.4 s) (Fig. 2F and SI Appendix, Fig. S2 E–H). Moreover, although beta–beta and delta–delta pairwise comparisons have an average delay near zero as expected, delta–delta pairwise comparisons have a much larger variance. This reflects a level of heterogeneity observed in delta cell Ca²⁺ behaviors that is lacking in beta cells within the same intact islets. Bootstrapping of the mean was calculated for each beta–delta cell pairwise comparison, one-sided P-values for all beta–delta cell pairwise comparisons in HG and 30 mM KCl were 0, indicating delta cells do precede beta cells (SI Appendix, Fig. S2 I and J).

Fig. 2.

Delta cell Ca²⁺ precedes beta cells. (A) Line graph of a single Ca²⁺ pulse in HG of a beta (green) and delta (red) cell overlaid (Top) with the slow (Middle) and fast (Bottom) components. Superimposed on the plot are symbols for the start of pulse (diamond), start of plateau (circle), end of plateau (triangle), and end of pulse (square). The maximum slope of the rise and fall of a calcium wave are also annotated as a dashed line over the slow component. (B) Top and Bottom are violin plots of the on rate and off rate respectively of beta and delta cells. (C) Histogram of the pairwise comparison of beta–beta, beta–delta, and delta–delta start of pulse times in HG. (D–F) The same line graph and histogram as A and B but in response to 30 mM KCl. Statistical significance was determined by the Mann–Whitney test (B and E) and Kruskal–Wallis test with Dunn’s multiple comparison test (C and F).

Slow Ca²⁺ Oscillations in Delta Cells Are Entrained by Beta Cells.

The synchronous Ca²⁺ activity of delta cells preceded that of beta cells in the same islet by many seconds, a longer delay than expected for a gap junction-mediated mechanism and contrary to the canonical model of pulsatile insulin secretion trailed by somatostatin secretion. To reconcile this unexpected observation, we set out to establish the interdependence of this oscillatory Ca²⁺ behavior of beta and delta cells. We crossed Ucn3-cre or Sst-cre to lsl-Salsa6f mice to selectively express both tdTomato and GCaMP6f in beta cells or delta cells, respectively; tdTomato fluorescence facilitated FACS purification of beta or delta cells. We then recorded the Ca²⁺ responses of individual FACS-purified beta cells or delta cells and compared this to their responses in intact islets from the same animals. Power spectral analysis of beta cell Ca²⁺ activity under HG stimulation within an intact islet demonstrated a narrow set of dominant frequencies corresponding to the well-established period of oscillation between 2 to 10 min (Fig. 3A) (33). The presence of multiple peaks in the spectral analysis reflects the intrinsic differences in oscillation frequency between individual islets imaged ex vivo. Likewise, power spectral analysis of the aggregated data from FACS-purified beta cells revealed a single wide peak at a slightly slower frequency of ~9 min, confirming that slow Ca²⁺ oscillations are generated cell-autonomously by each individual beta cell (Fig. 3B). Delta cells in intact islets exhibited a single dominant frequency between 5 to 10 min, mirroring the spectral analysis conducted in beta cells described above, along with a diffuse band of much faster frequencies reflecting beta cell-independent Ca²⁺ spikes (Fig. 3C). In contrast, while FACS-purified delta cells continued to demonstrate fast Ca²⁺ behaviors, they lost the slow oscillations present in delta cells in intact islets, as reflected by the absence of a peak in the power spectral analysis (Fig. 3D). These experiments demonstrate that slow, oscillatory delta cell Ca²⁺ responses are entrained by beta cells, despite preceding the response of neighboring beta cells by many seconds. Additionally, imaging beta and delta cell Ca²⁺ in alpha cell ablated islets demonstrates that the coordinated slow oscillations in beta and delta cells are independent of signaling from neighboring alpha cells (Fig. 3E).

Fig. 3.

Slow coordinated Ca²⁺ oscillations in delta cells are driven by beta cells. (A–D) Intensity plots of Ca²⁺ response (Left), Fourier transform of Ca²⁺ data (Right), and video stills of inactive (Bottom Left), active cells (Bottom Middle), and immunofluorescent post hoc stain (Bottom Right) from (A) beta cells in an intact Ucn3-cre x lsl-Salsa6f islet (Movie S4), (B) FACS-purified, dissociated Ucn3-cre x lsl-Salsa6f beta cells (Movie S5), (C) delta cells in an intact Sst-cre x lsl-Salsa6f islet (Movie S6), and (D) FACS-purified, dissociated Sst-cre x lsl-Salsa6f delta cells (Movie S7). (E) Intensity plot with accompanying video still (Top Right) and immunofluorescent stain (Bottom Right) of an islet isolated from a Gcg-cre x lsl-iDTR mouse with constitutive expression of GCaMP6s that permits the ablation of alpha cells upon administration of diphtheria toxin and allows calcium imaging in all islet endocrine cells (Movie S8).

Paracrine Signaling Is Necessary for Beta and Delta Cell Ca²⁺ Coordination in High Glucose.

The paradoxical observation that delta cell Ca²⁺ responses consistently precede beta cell Ca²⁺ responses by 16 s, while somatostatin secretion trails that of insulin by 30 s (10, 11), led us to investigate the possible contribution of a paracrine intermediate signal from the beta to the delta cell. To assay the role of paracrine signaling broadly, we applied the Rho-GTPase inhibitor ML-141 (34) to islets expressing GCaMP6s in beta and delta cells in HG. Rho-GTPase inhibition effectively inhibits exocytotic secretion from all islet cells (35, 36). Upon addition of ML-141, most delta cells that exhibited synchronous Ca²⁺ responses to neighboring beta cells lost this entrainment (Fig. 4A). This was evident by the decreased cross-correlation coefficient of the slow component between beta and delta cell Ca²⁺ activity in the presence of ML-141 when compared to an initial baseline and subsequent washout with HG (Fig. 4 B and C). Importantly, beta cell Ca²⁺ behavior was completely unaffected by ML-141 treatment (Fig. 4 A–D), and delta cell Ca²⁺ flux in the absence of beta cell activation was also unaffected (SI Appendix, Fig. S4 A–C). Since Ucn3 potentiates somatostatin secretion from delta cells and is coreleased with insulin by beta cells, we investigated whether Ucn3 may be the beta cell-derived paracrine factor that coordinates beta and delta cells. We first applied Astressin 2B (Ast2B), an inhibitor of the Ucn3 receptor Crhr2α that is expressed by delta cells (37), to islets expressing GCaMP6s in both beta and delta cells after stimulation with HG. Ast2B did not significantly decrease coordination between beta and delta cells (SI Appendix, Fig. S3 A, C, and D). Ucn3-knockout islets expressing GCaMP6s in both beta and delta cells modestly decreased coordination between beta and delta cells when compared to a heterozygous control, but this decrease was not statistically significant (SI Appendix, Fig. S3 B, E, and F). Because Ucn3 is copackaged with insulin, it likely stimulates cAMP from the G_αs-coupled Crhr2α receptor in an oscillatory manner (12). Indeed, when we imaged intracellular Ca²⁺ and cAMP dynamics simultaneously within the same islet, some delta cells exhibited oscillations in cAMP co-incident with changes in beta cell intracellular Ca²⁺, which would be in line with the scenario that paracrine factors such as Ucn3 secreted from beta cells promote delta cell cAMP oscillations (Fig. 4 E–G).

Fig. 4.

Many delta cells rely on paracrine signaling for coordinated Ca²⁺ oscillations in 16.8 mM glucose. (A) Intensity plot with line graphs overlaid for beta (green) and delta (red) cells. (B and C) Violin plot of Pearson’s correlation for (B) beta cells and (C) delta cells in relation to the islet average response in HG upon the addition of 10 µM ML-141 with highlighted insets in the intensity plot (A). (D) Image still (Left) and immunofluorescent stain (Right) of islet imaged in A from Movie S9. (E) Intensity plot of islet with both beta and delta cells expressing the adenovirus-delivered red Ca²⁺ reporter jRGECO in beta (green) and delta (red) cells, along with only delta cells expressing the genetically encoded FRET cAMP sensor, CAMPER (blue = low, yellow = high) with yellow line graph superimposed. (F) Magnification of the highlighted portion of E with delta cell Ca²⁺ removed. Vertical green shaded boxes highlight the regions of elevated beta cell Ca²⁺ during their response to HG. (G) Still image (Left) and immunofluorescent stain (Right) of islet depicted in E and F. Statistical significance was determined by one-way ANOVA with the Kruskal–Wallis test for multiple comparisons (B and C).

Many Delta Cells Are Gap Junction Coupled to Beta Cells Via Low-Density Connexin 36 Gap Junctions.

It has previously been concluded that beta and delta cells coordinate their activity via beta-like gap junctions. Nevertheless, the direct involvement of gap junctions and their potential identity has not been determined via delta cell-specific conditional deletion experiments. To this end, we referenced our previously published bulk RNA-seq data of FACS-purified delta cells (26) to identify the connexin genes expressed in delta cells. Connexin36 (Cx36) was the only detectable connexin in FACS-purified delta cells, albeit at 10-fold lower mRNA levels than in FACS-purified beta cells (Fig. 5A and SI Appendix, Fig. S5A) (26). Staining for Cx36 protein confirmed the presence of a punctate staining pattern, localized to the cell membranes of beta cells and some delta cells (Fig. 5B). To quantify the density and distribution of Cx36 gap junctions, we used lsl-mTmG transgenic mice, as the clear labeling of all membranes by tdTomato enabled the quantification of Cx36 puncta that coincide with plasma membranes. On average, delta cells (2.028 puncta/100 μm²) possessed fourfold fewer Cx36 puncta on their plasma membrane than beta cells (8.535 puncta/100 μm²) (Fig. 5C). The number of observed Cx36 puncta on beta cells was in line with published data from others (38). Additionally, 99% of all gap junctions detected on beta cells were present on a cell surface that contacted another beta cell; in contrast, 62.5% of delta cell gap junctions contacted another delta cell, while only 37.5% of gap junctions on delta cells contacted a neighboring beta cell (Fig. 5D). To rule out the contribution of connexins other than Cx36, we also stained for Cx43; we observed no staining within the islet, and prominent staining in the intercalated disks of cardiomyocytes where Cx43 is known to mediate synchronous cardiomyocyte contraction, validating our staining approach (SI Appendix, Fig. S5 B and C). To definitively establish the contribution of Cx36 gap junctions between beta and delta cells to the coordinated beta and delta Ca²⁺ responses in HG, we generated mice with a delta cell-specific knockout of Cx36 (Sst-cre x Cx36^fl/fl) (Fig. 5E). Using a combination of genetically encoded fluorescent calcium sensors (GCaMP6s and jRGECO1b) and the Ca²⁺ dye Calbryte 520, we measured changes in intracellular Ca²⁺ in both beta and delta cells. While these mice exhibit a small, but statistically significant decrease in beta cell coordination with respect to control mice, their beta cells continue to respond with synchronous Ca²⁺ oscillations (Fig. 5F). In contrast, delta cells without Cx36 exhibit a statistically significant and observable reduction in Pearson’s correlation in the slow component of their Ca²⁺ response, from a median of 0.7 in control mice to 0.3 in Sst-cre x Cx36^fl/fl mice (Fig. 5G).

Fig. 5.

Delta cells possess and require Cx36 gap junctions for coordinated Ca²⁺ oscillations in 16.8 mM glucose. (A) Bulk RNA sequencing data for Cx36, scaled from 0 to 50. (B) Immunofluorescent stain of lsl-mTmG pancreas for insulin (blue), Cx36 (green), membrane tdTomato (red), and somatostatin (white) (single channel panels in SI Appendix, Fig. S5D). (C) Bar graph of the number of Cx36 puncta/100 µm² of membrane for each individual beta and delta cell. (D) Proportion of beta and delta cells with gap junctions to neighboring beta (green) or delta (red) cells. (E) Intensity plot with line graphs overlaid of an Sst-cre x Cx36^fl/fl islet in response to HG of an islet expressing GCaMP6s (Movie S10). Ca²⁺ imaging was accomplished using GCaMP6s, Calbryte 520, and adenoviral transduction of JRGECO1b. (F and G) Violin plots of the Pearson’s correlation of each beta and delta cell relative to the islet average Ca²⁺ response of beta (green) and delta (red) cells from control islets (Figs. 1 and 2) and beta (blue) and delta (yellow) cells from Sst-cre x Cx36^fl/fl mouse islets. Statistical significance was determined using the Mann–Whitney test (C, F, and G).

Delta Cell Ca²⁺ Responses Are Heterogeneous and Correlate With Cx36 Density of Expression.

As we documented the Ca²⁺ behaviors of hundreds of delta cells, different patterns in their response to LG and HG emerged. To better document and classify this heterogeneity, we revisited the traces conducted in Figs. 2 and 3 of mouse islets expressing GCaMP6s in beta and delta cells. We were able to categorize delta cells based on two overarching features: their activity or inactivity in LG and the degree of coordination with beta cells in HG. Most, but not all delta cells were active under LG, with their activity in HG exhibiting a range of behaviors. Some delta cell Ca²⁺ behaviors closely mirrored the slow synchronous Ca²⁺ behaviors of beta cells with an almost complete absence of overlying fast Ca²⁺ responses, while others exhibited random-like fast Ca²⁺ spikes that lacked any observable coordination with beta cells. Based on the spectrum of Ca²⁺ behaviors observed under LG and HG stimulation, we subdivided the delta cell population into several groups (Fig. 6A). Of the 78 delta cells imaged from 4 mice, 77% exhibited some level of coordination with beta cells in high glucose with the remaining 23% exhibiting either uncoordinated Ca²⁺ spikes or no activity in HG at all. We characterized 37% of delta cells as “pseudosynchronous,” to reflect the fast uncoordinated Ca²⁺ spikes they exhibited during nadirs in the slow Ca²⁺ oscillations coordinated with beta cells. Approximately 27% of delta cells were inactive in low glucose and fully synchronous with beta cells in high glucose, a behavior most closely resembling the Ca²⁺ response of beta cells (Fig. 6A). As we only assayed delta cell activity in LG during a 20 to 30 min window, it is possible that some delta cells we characterized as silent under LG were in fact capable of responding but simply happened not to during the limited window of observation. When we imaged islets from Sst-cre x lsl-Salsa6f mice in LG for an extended 1-h period, 76% of delta cells exhibited at least a single Ca²⁺ spike. The remaining 24% of delta cells exhibited no Ca²⁺ activity (Fig. 6 B and C), confirming the presence of a population of delta cells that —like beta cells—is silent under LG.

Fig. 6.

Delta cells can be categorized based on their response to 2.8 mM and 16.8 mM glucose. (A) Intensity plot (Left) of delta cells, a portion of which were imaged and quantified in Fig. 2, categorized based on their respective behaviors, with their relative abundance as a percentage of the total delta cell population summarized (Right). (B) Intensity plot with overlaid line graphs of delta cells in response to 2.8 mM glucose alone. (C) Proportion of delta cells graphed against their average period of oscillation (red dots) with a curve fit using least square approximation to a hyperbolic function. (D) Graph of the proportion of beta and delta cells with detectable Cx36 gap junctions from the quantification of Cx36 immunofluorescent stain in Fig. 5. (E) Schematic of delta cell (red) crosstalk with neighboring beta cells (green) within an intact islet. One delta cell possesses Cx36 gap junctions (Left) and the other does not (Right).

To investigate whether this heterogeneity reflected differences in electrical coupling between beta cells and individual delta cells, we quantified the number of individual delta and beta cells with 0 or at least 1 punctum on their plasma membrane. Only 13% of delta cells had at least one detectable Cx36 punctum, compared to 43% of beta cells with at least one detectable Cx36 punctum (Fig. 6D). We can deduce from their synchronous Ca²⁺ behavior in intact islets that virtually all beta cells are gap-junction connected, with the under detection of Cx36 puncta in each beta cell caused by the fact that we imaged and analyzed only a relatively thin optical slice of 528 nm in depth, corresponding to the limit of our axial resolution. Nevertheless, this degree of underdetection of actual Cx36 gap junctions would apply equally to beta and delta cells. We therefore conclude that—owing to the overall fourfold lower average density of Cx36 gap junctions in delta cells we observed—it is likely that a subset of delta cells possesses no Cx36 gap junctions, contributing to the inflated number of delta cells observed to have no Cx36 gap junctions in our data. This would be fully in line with our observation that 23% of all delta cells fail to demonstrate a meaningful coordination with the Ca²⁺ behavior of neighboring beta cells. Conversely, the delta cells that are most beta cell-like based on the combination of their lack of Ca²⁺ under LG and tight synchronization under HG are likely those that possess the highest density of Cx36 gap junctions with neighboring beta cells (Fig. 6E). To test this notion, we revisited and stratified the Ca²⁺ response times to HG described in Fig. 3 based on whether the observed delta cells were “synchronous” or “pseudosynchronous” with neighboring beta cells. For three of the four parameters measured (start of pulse, start of plateau, and end of plateau), the delta cells most synchronous to beta cells also demonstrated significantly shorter offsets in response relative to beta cells compared to pseudosynchronous delta cells (SI Appendix, Fig. S6 A–D).

Discussion

In this study, we provide a comprehensive assessment of the signaling mechanisms that facilitate beta and delta cell crosstalk and demonstrate the requirement for both paracrine signaling and gap junction coupling. Using mice expressing GCaMP6s in both beta and delta cells, we assessed the similarities and considerable differences in their Ca²⁺ behaviors. We find that the Ca²⁺ responses of most—but not all—delta cells become entrained by beta cells upon HG stimulation. Paradoxically, delta cell Ca²⁺ responses precede beta cell Ca²⁺ responses in HG and KCl by an average of 16 s. Moreover, a large majority (76%) of delta cells demonstrate uncoordinated fast Ca²⁺ spikes under LG, when beta cells are silent. We demonstrated by immunofluorescence and delta cell-specific Cx36 knockout that gap junctions containing Cx36 between a portion of delta cells and their neighboring beta cells are required for beta cell entrainment of delta cell Ca²⁺ responses. Nevertheless, the density of Cx36 on heterotypic beta–delta connections is significantly lower compared to beta–beta homotypic interactions, likely explaining why delta cells can engage in Ca²⁺ activity relatively unimpeded when beta cells are collectively silent. In addition to gap junctions, we demonstrate that paracrine signals are also required—as acute inhibition of exocytosis selectively perturbs beta cell-entrained Ca²⁺ oscillations in delta cells as well. Taken together, our findings explain the coordination between beta and delta cells via a combination of paracrine crosstalk and relatively weak electrical connections that facilitate coordination, without constraining the considerable individuality of delta cell Ca²⁺ behaviors.

Cx36 Gap Junctions Play Very Different Roles in Beta and Delta Cells.

In work published several years ago, respected colleagues in our field proposed that delta cells are coupled to beta cells via beta cell-like gap junctions (25). This conclusion was based on experiments in which optogenetic activation of channelrhodopsin-2 selectively in beta cells was observed to rapidly activate neighboring delta cells. The presence of gap junctions between beta and delta cells was tested by the application of the gap-junction inhibitor carbenoxolone, which interrupted glucose-stimulated inward current in delta cells. However, carbenoxolone would also have inhibited gap junction connections between beta cells, and any downstream paracrine signal to delta cells. While the involvement of Cx36 in forming gap junctions between beta and delta cells was implied, this was not experimentally validated at the time. The presence of gap junctions between beta and delta cells offers a conceptually straightforward hierarchical model of beta cell control over delta cell activity. However, as we demonstrate here, plain observation of delta cell Ca²⁺ behavior under glucose concentrations below the beta cell glucose threshold makes it clear that most delta cells demonstrate Ca²⁺ behaviors independently from neighboring beta cells. We and others have shown that this Ca²⁺ behavior and somatostatin secretion depend on the closure of K_ATP channels and opening of L-type voltage gated Ca²⁺ channels (SI Appendix, Fig. S1) (28, 39, 40). The ability of delta cells to depolarize independently contrasts directly to beta cells, which are well established to be connected by Cx36 gap junctions that turn the entire beta cell mass of an islet into an electrically coupled functional syncytium. This arrangement not only ensures the synchronous response of all beta cells in an islet to HG stimulation, but—crucially—helps ensure beta cell inactivity below the beta cell glucose threshold of 6 mM glucose (15, 21, 41, 42).

Yet, while all beta cells within an islet respond in quick succession of each other, delta cells in the same islet respond to the same stimuli on average 16 s before beta cells—an eternity when considering the propagation of ions and small molecules via gap junctions. These behaviors are not reflective of what one might expect from beta and delta cells that are connected by gap junctions. Yet, using a combination of immunofluorescence and delta cell-specific deletion of Cx36, we provide here the definitive demonstration that delta and beta cells share relatively low numbers of gap junctions that contain Cx36. Crucially, we quantified that average Cx36 gap junction density on beta–delta heterotypic contacts are fourfold lower compared to beta–beta homotypic membranes, in good general agreement with the 10-fold lower mRNA expression of Cx36 (26) and 10-fold lower gap junctional conductance formed between beta and delta cells relative to those between two beta cells (19). This low gap junction density is key as it may explain how delta cells are capable of entrainment by electric signals from beta cells without being clamped inextricably to the beta cell membrane potential as a higher density of gap junctions would do. It is likely that the considerable heterogeneity we observe correlates directly with Cx36 gap junction density. The subset of delta cells that demonstrates Ca²⁺ behaviors that are most beta cell-like likely share the highest density of gap junctions. Conversely, the 21% of delta cells that are fully uncoordinated with beta cell in HG likely lack gap junctions—in line with the lack of detectable Cx36 on most delta cells. These gap junction-negative delta cells may have eluded investigators in earlier studies (25), as they specifically excluded delta cells active in LG, a category that encompasses as many as 76% of all delta cells.

Paracrine Signaling From Beta Cells Is Necessary for Coordination Between Beta and Delta Cells.

Our data also support an important role for paracrine signaling in the coordination of beta and delta cell Ca²⁺ dynamics. While the identity of this paracrine factor remains unknown, it is clear based on the loss of coordination observed upon the addition of ML-141 that inhibition of beta cell secretion interrupts Ca²⁺ entrainment by beta cells of delta cells. Our first inclination was to test the delta cell activator Ucn3; however, deletion of Ucn3 or pharmacological blockade of its receptor had limited direct effect on coordinated Ca²⁺ oscillations between beta and delta cells. Despite this, Ucn3 is a robust stimulator of somatostatin secretion and thus may provide an explanation for the discrepancy between the early rise in delta cell Ca²⁺ and late secretion of somatostatin with respect to beta cell Ca²⁺ oscillations and insulin secretion. This exemplifies an important distinction between Ca²⁺ and secretion—while depolarization-dependent Ca²⁺ is necessary for secretion, on its own it may not promote robust secretion (29).

Integrated Model of Beta and Delta Cell Coordinated Signaling in High Glucose.

We therefore propose a model for coordinated beta and delta cell Ca²⁺ dynamics and secretion whereby beta and delta cells coordinate their Ca²⁺ dynamics via a combination of Cx36-mediated gap junction coupling and paracrine signaling from neighboring beta cells (Fig. 6E). The enhanced excitability of the delta cell, along with the lower gap junctional conductance and presence of a paracrine factor acting to modulate Ca²⁺ activity, leads to early delta cell activation. This heightened delta cell excitability is exemplified by both the faster on/off rate and earlier response time of delta cells compared to beta cells in both HG and KCl. Indeed, delta cells exhibit a much higher membrane resistance due to a smaller pool of open K_ATP channels at rest when compared to beta cells (39). This indicates that delta cells require a proportionally smaller current to initiate a similar change in membrane potential than beta cells. Conversely, the strong electrical coupling between beta cells may also lengthen the time from initial K_ATP channel closure and graded increase in membrane potential (which is the event that would trigger entrainment of delta cells) to subsequent flux of Ca²⁺ in beta cells. As delta cells exhibit a much higher membrane resistance, they likely respond faster to any small, graded change in beta cell membrane potential than neighboring beta cells do, which manifests as delta cell Ca²⁺ preceding beta cell Ca²⁺. This early activation of Ca²⁺ serves to prime the delta cell, but importantly does not itself stimulate robust somatostatin secretion, as it is established that this requires cAMP (29), which within the islet is provided by signaling from the beta cells, like Ucn3 (12). This would explain the 30 s delay between synchronous delta cell Ca²⁺ activation and synchronous somatostatin secretion that has been reported by others (10, 11). Such a delay would be in good agreement with the time it would take for beta cell paracrine factors such as Ucn3 to act on delta cells to amplify pulsatile somatostatin secretion triggered by beta cell entrainment of delta cells—a sequence of events that is supported by simultaneous Ca²⁺ and cAMP imaging in delta and beta cells (Fig. 4 E–G).

In summary, we have demonstrated that delta cells communicate with beta cells via a combination of gap junction coupling and paracrine signaling. The exact mechanism and relative contribution of these mechanisms varies among the delta cell population, resulting in a spectrum of heterogeneous individual delta cell behaviors in response to changes in glucose concentrations. Additionally, we propose a comprehensive model of coordinated somatostatin release whereby the initial rise in Ca²⁺ observed in delta cells acts to prime delta cells but does not fully stimulate somatostatin secretion until beta cell activation and the resulting release of paracrine delta cell activators. Our observations demonstrate how different densities of the same gap junction expressed by neighboring cell populations support fundamentally different Ca²⁺ behaviors, leading to notably heterogenous delta cell responses and uniform beta cell behaviors. It is well established that exogenous somatostatin can help diabetic patients maintain euglycemia with less insulin, thereby reducing the risk of iatrogenic hypoglycemia and improving glycemic control (43). Its use however is limited by its short half-life and off-target effects (43). Thus, the selective modulation of delta cell secretion offers a source of somatostatin that can act locally within the islets of Langerhans, obviating the off-target effects of similar pharmacologic approaches. By understanding the mechanism by which beta cells normally modulate delta cell somatostatin secretion, we may one day be able to develop novel therapeutics capable of selectively targeting delta cells to improve the glycemic control of diabetic patients and those with excessive insulin secretion, with fewer side effects than current options.

Limitations of This Study.

Disentangling the effects of paracrine signaling from that of gap junction coupling from beta cells onto delta cells was a major challenge in this study. Traditional methods of gap junction inhibition such as 18α-glycyrrhetinic acid, mefloquine, or carbenoxolone affect both beta–beta cell coupling and beta–delta cell coupling. Simply inhibiting gap junctions within the islet without a more targeted approach would invariably lead to changes in paracrine signaling from beta cells, creating an insurmountable confound. For this reason, we relied on the data collected from conditionally floxed mice enabling the selective deletion of Cx36 in delta cells to establish the role of gap junction coupling. Likewise, suppressing secretion without directly inhibiting or altering Ca²⁺ signaling was a major challenge as well. For this reason, we used data collected from islets exposed to the Cdc42 inhibitor ML-141, which is to the best of our knowledge the only compound that can suppress secretion without suppressing Ca²⁺. We acknowledge that more targeted approaches such as electrophysiology offer superior resolution at the single cell level but argue that the observations made with respect to the heterogeneity of delta cells was made possible only via the considerably higher throughput offered by live cell imaging of assemblies of beta and delta cells within the same islets. Islet isolation and maintenance in culture for 1 to 2 d before imaging also introduces potential confounds that may limit the generalizability of this study to the true in vivo activity of islets. However, imaging islets in vivo is technically challenging and precludes the use of drugs like ML-141, diazoxide, and isradipine, as their introduction into a living mouse would themselves have off-target effects on other organ systems that may then indirectly affect islet function. Last, the use of a single fluorescent calcium sensor in both beta and delta cells is limited by the potential for changes in fluorescent intensity in one cell to influence the measured change in fluorescence in adjacent cells. This potential confound was limited by the use of the Fiji plugin “image stabilization,” which reduces any lateral movement of the islet during imaging, and drawing the ROIs smaller than the full size of an individual cell and manually verifying each video with the ROIs superimposed to greatly reduce a scenario where an individual ROI inadvertently captures activity of a neighboring cell.

Materials and Methods

Animals.

Mice were kept on a 12-h light:dark cycle with free access to water and standard chow. Unless otherwise stated, mice used were between 3 to 6 mo old. All experiments were approved by the UC Davis Institutional Animals Care and Use Committee. All mice used were on the C57BL/6 background, with the exception of the Cx36^fl/fl mice, which were originally on an agouti background and crossed to mice on the C57BL/6 background for 2 generations. Mouse generation and genotypes are in the extended materials and methods.

Islet Isolation.

Islet isolation procedure has been previously described (15) and is outlined in the extended materials and methods. Islet dissociation was accomplished via a 2 min incubation in 1 mL of 0.25% Trypsin (Gibco cat:25200056) at 37ºC, then mechanically dissociated via repeated pipetting. Once islets were no longer visible in the dissociated islet suspension, 9 mL of complete RPMI was added and the dissociated cells were pelleted via centrifugation at 1,000 RPM for 10 min. The supernatant was decanted and the pellet resuspended in 1 mL of complete RPMI. FACS purification was conducted by the UC Davis Flow Cytometry Core using a 100 μm wide nozzle. Salsa6f tdtomato-positive cells were collected in a 1:1 mixture of complete RPMI and FBS, centrifuged at 5,000 RPM for 5 min, resuspended in complete RPMI, and plated in microfluidic chambers at 10,000 to 20,000 cells/chamber.

Live Cell Ca²⁺ Imaging.

Live cell imaging was conducted on islets as previously described (15), with detailed information in extended materials and methods. Krebs Ringer Buffer (KRB) was used as the perfusion media, with either 2.8 mM, 5.5 mM, or 16.8 mM glucose added. All traces begin with a 10-min washout of 2.8 mM glucose KRB to wash away residual RPMI.

Sst-cre × Cx36^fl/fl islets were imaged using a combination of genetically encoded GCaMP6s (lsl-GCaMP6s), adenovirus transduction of jRGECO1b, and Calbryte 520 (AAT Bioquest cat:20650). Adenovirus transduction was accomplished by incubating cells overnight with a 1:1,000 dilution of jRGECO1b virus. The virus was then subsequently washed out with RPMI and plated into microfluidic chambers bound to Mattek glass bottom dishes. Loading of islets with Calbryte 520 was accomplished via incubation of 5.5 mM glucose in KRB with Calbryte 520 and pluronic F-127 (Sigma Aldrich cat:P2443-250G) for 1 h, after which islets within microfluidic chambers were placed on the microscope and perfused to wash away any residual Calbryte 520. Each of three Sst-cre x Cx36^fl/fl mice were imaged using either GCaMP6s, Calbryte 520, or adenoviral transduction with JRGECO1b. The results were similar irrespective of calcium indicator and showed oscillating beta cell calcium responses that were no longer entraining delta cells. We therefore pooled these data for comparison to the control islets with wild-type Cx36 expression from mIP-H2B-Chy x Sst-cre x Ucn3-cre x lsl-GCaMP6s reported mice used in Figs. 1 and 2.

Ca²⁺ Image Analysis.

Regions of interest (ROIs) were drawn manually using either NIS Elements or Fiji and were drawn well within the margins of the cell of interest. For videos with mild lateral movement, the “image stabilization” plugin in Fiji was used. Average fluorescence intensity within each ROI at each time frame was exported as a .csv file along with the area, in μm, of each ROI. A detailed description of the custom Python script used to filter the slow and fast components of each Ca²⁺ trace is available in the extended materials and methods section. To calculate the Pearson correlation values of individual cells, the slow component of Ca²⁺ activity within an islet was averaged and normalized. The Pearson correlation value was then calculated between each cell’s normalized slow component Ca²⁺ response and the islet average response.

Immunofluorescence.

After imaging, chambers were fixed and immunostained as described previously (15). Immunofluorescence of sectioned pancreas for the detection of Cx36 was achieved via transcardial perfusion of deeply anesthetized mice. Detailed procedures are available in the extended materials and methods.

Cx36 Immunofluorescence Analysis.

Images from mTmG mice stained for Cx36, insulin, and somatostatin were first segmented using a custom-trained model in Cellpose 3.1.1.1. Cx36 puncta were identified using the segmentation protocol outlined (44). Identification of beta and delta cells was done semiautomatically using a custom python script described in the extended materials and methods with a final pass and update conducted by a human.

Statistical analysis.

Statistical tests were computed using Prism GraphPad Software (10.3.1) with normality and lognormality tests (D’Agostino and Person test, Anderson–Darling test, Shapiro–Wilk test, Kolmogorov–Smirnov test, and QQ plot) run on each dataset with the appropriate statistical test applied based on the normality or lack thereof. Statistical tests for each experiment are detailed in the figure legend with a P-value < 0.05 being considered significant. Bootstrapping was accomplished in Python via random sampling with replacement of each beta–delta pairwise comparison dataset. The mean of each random sample was calculated, and the random sampling was repeated 10,000 times. One-tailed P-values were calculated by dividing the sum of bootstrapped means less than or equal to zero by 10,000.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Movie S1.

Video of islet from a mouse expressing GCaMP6s in delta cells only. Islet is exposed to 5.5 mM glucose for 130 minutes with the addition of 100 μM diazoxide and 2 μM isradipine for 15 minutes each and ending with 30 mM KCl for 2 minutes. Accompanying the video is audio compiled from the average GCaMP6s response. Accompanies Figure S1D.

Movie S2.

Video of islet expressing GCaMP6s in beta and delta cells. Islet is exposed to 5.5 mM glucose for 130 minutes with the addition of 100 μM diazoxide and 2 μM isradipine for 15 minutes each and ending with 30 mM KCl. Accompanying the video is audio compiled from the average GCaMP6s response with beta cells being represented by the lower frequency noise and delta cells by the higher frequency noise. Accompanies Figure S1H.

Movie S3.

Video of islet expressing GCaMP6s in beta and delta cells. Islet is exposed to 16.8 mM glucose for 1 hour and then 2.8 mM glucose for 45 minutes. During the 2.8 mM glucose washout 100 nM ghrelin is added for 10 minutes with the trace ending in 30 mM KCl. Accompanying the video is audio compiled from the average GCaMP6s response with beta cells being represented by the lower frequency noise and delta cells by the higher frequency noise. Accompanies Figure 1A.

Movie S4.

Video of islet expressing GCaMP6s and tdTomato in beta cells only. Islet is first exposed to 2.8 mM glucose for 10 minutes then 16.8 mM glucose for 1 hour and lastly 2.8 mM glucose for 40 minutes. During the 2.8 mM glucose washout 100 nM ghrelin is added for 10 minutes with the trace ending in 30 mM KCl. Accompanying the video is audio compiled from the average GCaMP6s response. Accompanies Figure 3A.

Movie S5.

Video of individual FACS purified beta cells expressing GCaMP6s and tdTomato in. Islet is first exposed to 2.8 mM glucose for 10 minutes then 16.8 mM glucose for 1 hour and lastly 2.8 mM glucose for 40 minutes. During the 2.8 mM glucose washout 100 nM ghrelin is added for 10 minutes with the trace ending in 30 mM KCl. Accompanying the video is audio compiled from the average GCaMP6s response. Accompanies Figure 3B.

Movie S6.

Video of islet expressing GCaMP6s and tdTomato in delta cells only. Islet is first exposed to 2.8 mM glucose for 10 minutes then 16.8 mM glucose for 1 hour and lastly 2.8 mM glucose for 40 minutes. During the 2.8 mM glucose washout 100 nM ghrelin is added for 10 minutes with the trace ending in 30 mM KCl. Accompanying the video is audio compiled from the average GCaMP6s response. Accompanies Figure 3C.

Movie S7.

Video of individual FACS purified delta cells expressing GCaMP6s and tdTomato. Islet is first exposed to 2.8 mM glucose for 10 minutes then 16.8 mM glucose for 1 hour and lastly 2.8 mM glucose for 40 minutes. During the 2.8 mM glucose washout 100 nM ghrelin is added for 10 minutes with the trace ending in 30 mM KCl. Accompanying the video is audio compiled from the average GCaMP6s response. Accompanies Figure 3D.

Movie S8.

Video of islet expressing Gcg-cre x lsl-iDTR x con-GCaMP6s where in alpha cells have been ablated via administration of diphtheria toxin in vivo prior to islet isolation. Islet is first exposed to 2.8 mM glucose for 10 minutes then 16.8 mM glucose for 1 hour and lastly 2.8 mM glucose for 40 minutes. During the 2.8 mM glucose washout, 100 nM ghrelin is added for 5 minutes and 100 nM epinephrine is added for 5 minutes with the trace ending in 30 mM KCl. Accompanies Figure 3E.

Movie S9.

Video of islet expressing GCaMP6s in beta and delta cells. Islet is first exposed to 2.8 mM glucose for 10 minutes then 16.8 mM glucose for 90 minutes with 10 μM ML-141 added in the middle for 30 minutes and lastly 2.8 mM glucose for 35 minutes. During the 2.8 mM glucose washout 100 nM ghrelin is added for 10 minutes with the trace ending in 30 mM KCl. Accompanying the video is audio compiled from the average GCaMP6s response with beta cells being represented by the lower frequency noise and delta cells by the higher frequency noise. Accompanies Figure 4D.

Movie S10.

Video of islet expressing GCaMP6s in beta and delta cells. Islet is first exposed to 2.8 mM glucose for 10 minutes then 16.8 mM glucose for 1 hour and lastly 2.8 mM glucose for 35 minutes. During the 2.8 mM glucose washout 100 nM ghrelin is added for 10 minutes with the trace ending in 30 mM KCl. Accompanying the video is audio compiled from the average GCaMP6s response with beta cells being represented by the lower frequency noise and delta cells by the higher frequency noise. Accompanies Figure 5E.

Movie S11.

Video of islet expressing GCaMP6s in all islet cells exposed to 5.5 mM glucose with a 30-minute pulse of 10 μM ML-141, 10 minutes of 100 nM ghrelin, and 30 mM KCl at the end of the trace. Accompanies Figure S4B.

Data, Materials, and Software Availability

Scripts used for this publication are freely available to download using our GitHub repository: https://github.com/Huising-Lab (45). All study data are included in the article and/or supporting information.