Dynamic Uni- and Multicellular Patterns Encode Biphasic Activity in Pancreatic Islets

Biphasic secretion is a common physiological feature in a number of hormone and neuro-hormone–secreting micro-organs (1–4). Pancreatic islets represent a well-described model of biphasic secretion (4,5): a first peak phase (5–15 min) is followed by a decrease in the secretion rate, called nadir, and a subsequent second long-lasting plateau phase (6,7) and installation of pulsatility (8). Insulin secreted during the first phase immediately reaches the liver to rapidly regulate blood glucose levels. The second phase targets more distant organs as long as glycemia remains elevated (9). This optimized kinetic is strongly altered during aging and in type 2 diabetes (10–14). As biphasic insulin profiles persist ex vivo (6), multiorgan feedback loops are not required, and patterns are encoded at the micro-organ level.

Although the phenomenon per se has been extensively described, it is still not understood how this phasic organization is achieved and what drives the progression from first to second phase. β-Cell metabolism has been monitored via mitochondrial membrane potential, oxygen consumption, or metabolic coupling factors. Metabolism increases upon glucose stimulation, with, often (15,16) but not always (17,18), a discrete and brief peak of 1–2 min during the first phase before raising again during the nadir while secretion decreases; thus, these metabolic profiles do not explain secretion patterns. The organization of insulin-secreting vesicles in distinct functional pools in β-cells has been widely invoked to explain biphasic secretion (19–21). Interestingly, biphasic activation is a multicellular process since it is profoundly altered in dissociated islets and connexin-36 knockout mice (22,23). Hence, vesicle pool organization may not represent the main determinator of biphasic activation in islets.

Different approaches have been used to investigate multicellular processes in islets. Analysis of intraislet connectivity by dynamic imaging has provided an elegant model of highly active leader cells (24,25), in line with β-cell heterogeneity (26), but requires complex mathematical offline reconstruction and may potentially introduce bias (24,27). In addition, the existence of such hub cells is still debated (28), and inherent restrictions have limited such experiments to short periods. Consequently, they do not inform about the dynamic evolution of the entire micro-organ.

We have therefore sought for a more direct approach endowed with high temporal resolution (kHz) and useable throughout the hours of postprandial islet activation, a situation in which rundown in optical and classical electrophysiological approaches may occur. Analysis of extracellular electrical field potentials with multielectrode arrays (MEAs) of intact islets avoids such drawbacks (29,30). Both unicellular and multicellular signals can be observed in the form of single-cell action potentials (APs) (29,31) and multicellular slow potentials (SPs) for hours or even days (30,32). SPs represent robust and specific signals propagated among β-cells via connexin-36 in both rodent and human islets (32). Hence, this approach provides a dynamic, direct, and unbiased measurement of unicellular and of micro-organ behavior via APs and SPs.

We have therefore addressed the question how biphasic activity of pancreatic islet micro-organs is encoded in terms of single-cell and coupled electrical activity throughout a physiological time span and how this is disrupted in pathophysiological states.

Adult male C57BL/6J mice (10–20 weeks of age, except for Fig. 7A and B: 12–45 weeks) were sacrificed by cervical dislocation according to University of Bordeaux ethics committee guidelines. Islets were obtained by enzymatic digestion and handpicking (29,30,32). MEAs were coated with Matrigel (2% v/v) (BD Biosciences, San Diego, CA), and intact islets were seeded (one pancreas per MEA) and cultured at 37°C (5% CO₂, >90% relative humidity) in RPMI medium (11 mmol/L glucose, except for glucotoxic conditions: 20 mmol/L in Fig. 7C and D) (Thermo Fisher Scientific, Waltham, MA) as described (29,30,32).

Human islets (healthy donors; for details, see Supplementary Material) were isolated at the Geneva Cell Isolation and Transplantation Center (29,32), distributed through the European Consortium for Islet Transplantation, and authorized by the ethical committee (Comités de Protection des Personnes; 16-RNI-10). Human islets were cultured on MEAs under the same conditions as mouse islets but using CMRL-1066 medium (5.6 mmol/L glucose, except for glucotoxic conditions) (Thermo Fisher Scientific) (29,32).

Different MEAs (Multi Channel Systems GmbH [MCS], Reuttlingen, Germany) were used to address specific questions. Standard MEAs (60MEA200/30iR-Ti-gr, 59 titanium nitride electrodes [TiN], Ø 30 µm, 200 μm interelectrode distance) permitted the recording of SPs of ∼1 islet/electrode (1.0 ± 0.1 islets/electrode, n = 49 islets, N = 3 independent preparations). Recordings of different intraislet regions were performed using high-density MEAs (HD-MEAs) (60HexaMEA40/10iR-ITO-gr, 59 TiN electrodes, Ø 10 µm, 40 μm interelectrode due to the flow). Both MEAs were continuously perifused at 0.5 mL/min (Reglo ICC; Ismatec, Glattbrugg, Switzerland).

To measure simultaneously electrical parameters and insulin secretion (ELISA 80-INSMSU-E01; ALPCO, Salem, NH), microfluidic MEAs (µMEAs) were developed using MEA200/30iR-Ti-gr with a microfluidic channel (Ø 0.8 mm) in polydimethylsiloxane and perfused at 8 μL/min (MFCS-EZ; Fluigent, Villejuif, France). Kinetics of medium changes were determined as published (30).

Finally, poly(3,4-ethylenedioxythiophene) (Pedot) and carbon nanotube–covered MEAs (Pedot-MEAs) (60PedotMEA200/30iR-Au-gr, electrode arrangement as in standard MEAs) were used to detect APs, which are hardly discernable otherwise (29), and solutions were replaced by pipetting to avoid mechanical artifacts.

MEA recordings were performed at 37°C, pH 7.4 (29,32), in solutions containing 1.2 mmol/L CaCl₂ for mouse islets (2.5 mmol/L in Supplementary Fig. 5B–D) or 1.8 mmol/L for human islets as published previously (29,32), which is close to physiological levels and provides sufficient driving force for SP quantification. When specified, a solution without CaCl₂ was applied to evaluate basal activity. Glucagon-like peptide 1 (GLP-1) solutions (Bachem Bio-Science Inc, King of Prussia, PA) were prepared ex tempore. Electrodes with noise levels >30 μV peak-to-peak were regarded as artifacts, connected to the ground, and not analyzed (3.6 ± 1.7% of electrodes; N = 5). Extracellular field potentials were acquired at 10 kHz, amplified, and band-pass filtered at 0.1–3,000 Hz with a USB-MEA60-Inv-System-E amplifier (MCS; gain: 1,200) or an MEA1060-Inv-BC-Standard amplifier (MCS; gain: 1,100), both controlled by MC_Rack software (v4.6.2, MCS).

Intracellular recordings were performed simultaneously with extracellular recordings on standard MEAs. Intracellular potentials of islet cells were measured by current-clamp (10 kHz sampling rate, 10 kHz low-pass filter) with sharp glass Clark micropipette microelectrodes (Harvard Apparatus, Les Ulis, France) filled with 3 mmol/L KCl (96 ± 18 MegaΩ; n = 17) and coupled to an Axoclamp-2B amplifier (Molecular Devices, San Jose, CA) controlled by Spike2 software (v7.01; Cambridge Electronic Design Ltd, Cambridge, U.K.). A common reference electrode was used for both recordings. An electrical artifact observable on both recordings was used to synchronize intra- and extracellular traces.

Images of islets on MEAs were taken before and after each experiment to localize electrodes covered with islets (44.1 ± 7.4% of electrodes; N = 5 independent preparations). Islet cell monolayer surfaces were quantified with ImageJ software (v1.52d; National Institutes of Health, Bethesda, MD). Electrophysiological data were analyzed with MC_Rack software. SPs and APs were isolated using a 2-Hz low-pass filter or a 3–700-Hz band-pass filter, respectively, and frequencies were determined using the threshold module of MC_Rack with a dead time (minimal period between two events) of 300 ms (SPs) and 10 ms (APs). The peak-to-peak amplitude module of MC_Rack was used to determine SP amplitudes. Simultaneous extra- and intracellular recordings were analyzed with Spike2 software.

After filtering at 2 Hz, SPs were detected using the peak detection module of Spike2 with a threshold of −15 μV. The degree of synchrony between SPs on electrodes was computed with MATLAB (vR2012B; MathWorks, Natick, MA) following a method based on Schreiber et al. (33), originally used to compute synchronization between trains of neuronal spikes (Supplementary Methods and Supplementary Fig. 1).

Experiments were replicated on at least three independent biological preparations, except when indicated. If not stated otherwise, N represents the number of independent preparations and n the number of electrodes analyzed. Graphics, quantifications, and statistics were performed with Prism software (v7; GraphPad, La Jolla, CA). Data are presented as means and SEM or whiskers boxes (box, 25th to 75th percentiles; line in the middle of the box, median; + or ■, mean; and whiskers, 10–90th percentiles). The minimal value of mean SP frequency after the first peak (corresponding to the nadir) was taken as the limit between phases.

Gaussian distributions were tested by D’Agostino-Pearson test and comparison of two groups with paired data by two-tailed paired t tests or nonparametric t tests with Wilcoxon matched-pairs signed-rank test. Two groups with unpaired data were compared using two-tailed unpaired t tests or nonparametric Mann-Whitney tests. For more than two groups, one-way ANOVA with Tukey post hoc or nonparametric Dunn tests were used.

Data sets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Intact mouse islets were cultured on MEAs (Fig. 1A) to record noninvasively extracellular field potentials. As previously described, two types of signals were observed (Fig. 1B): the well-known unicellular APs and the more recently described SPs (29,30), which are of multicellular origin and require β-cell coupling via connexin-36 (32). We first monitored the frequency and amplitude of SPs in islets to determine whether they correlate to well-known biphasic secretion patterns. When islets were stimulated by an increase in glucose from 3 mmol/L to the moderate concentration of 8.2 mmol/L (Fig. 1C), a clear biphasic electrical profile of SPs was triggered, in terms of both frequencies and amplitudes (Fig. 1C). Each phase owned a specific electrical “signature”: SPs of high frequencies but small amplitudes in the first phase and lower frequencies but increasing amplitudes during the second phase (Fig. 1D). Hence, electrical coupling modes of islet β-cells are biphasic and develop in a dynamic fashion.

Another known insulin secretagogue, l-leucine, bypasses glycolysis. Stimulating the same islets with either glucose or leucine (Fig. 1E), the first electrical phase was comparable between the two molecules in terms of peak of SP frequencies. However, in the case of leucine, SPs were largely reduced in the second phase (Fig. 1E and F). Thus, the metabolism of the main stimulator (i.e., glucose) triggers a full second electrical phase, while leucine may require coactivation of additional metabolic pathways, such as by glutamine (34).

By introducing microfluidics in µMEAs (Fig. 2A and B), we simultaneously recorded SPs and insulin secretion (Fig. 2C). Biphasic kinetics of SP frequencies were highly correlated with biphasic insulin secretion (Fig. 2C). Moreover, maximal correlation was obtained when both SP frequencies and amplitudes were taken into account (Fig. 2C and D), supporting the view that multicellular SPs, upstream of secretory pools, constitute the main regulator of biphasic insulin secretion.

The magnitude of extracellular electrical signals often mirrors the degree of cell synchrony, at least in the brain (35). We hypothesized that the increase of SP amplitude during the second phase was due to an increase in β-cell synchrony within an islet. Since standard MEAs do not offer spatial resolution below the dimensions of a single islet (Supplementary Fig. 2A), we used HD-MEAs providing ∼10 times more electrodes per islet, which permitted multisite analysis of single islets without affecting the signal-to-noise ratio (Supplementary Fig. 2B). The two electrical phases of SPs were again clearly observable regarding frequencies and amplitudes (Supplementary Fig. 2C and D) and pulsatility appeared after 40–60 min with a mean pulse period of ∼4 min (Supplementary Fig. 2D–F), similar to insulin secretion data (8,36). Note that islet inactivation following glucose decrease was rapid (<5 min) and not phasic (Supplementary Fig. 2C and D). In the representative HD-MEA illustrated in Fig. 3, two electrodes covered by the same islet were compared with each other (intraislet) and to a third electrode contacting another islet (interislets) (Fig. 3A). All three electrodes revealed SPs of high frequencies and small amplitudes during the first phase and the inverse during the second phase (Fig. 3B). The increase in SP amplitudes in the second phase was concomitant with a clear synchronization of SPs within a given islet, but not between different islets (Fig. 3B). Such absence of interislets synchrony suggests that extraislet mechanisms are involved in the whole pancreas synchrony during pulsatile secretion (8,36). A dynamic MATLAB code was then developed to quantify the degree of interelectrode synchrony during the biphasic activation (Supplementary Methods and Supplementary Fig. 1). At 3 mmol/L glucose, islets rarely generated SPs, and consequently, intraislet synchronies were absent (Fig. 3C, left). When islets were stimulated by 8.2 mmol/L glucose, the different regions within the same islet partially synchronized in the first phase (Fig. 3C, middle), and this intraislet synchrony considerably increased during the second phase (Fig. 3C, right). Statistical comparisons confirmed specific electrical coupling modes for each phase, and SP synchrony was positively correlated with amplitude and negatively correlated with frequency (Fig. 3D). Thus, SP amplitude represents a direct, nonbiased, and continuous measurement of intraislet connectivity, and synchrony is accompanied by the generation of larger functional clusters of lesser activity.

We further addressed the nature of the SPs and islet β-cell coupling by simultaneous extracellular and intracellular recordings. To that end, standard MEAs were used: SPs were measured on a microelectrode at steady state during the second phase, while membrane potentials were recorded with sharp microelectrodes introduced into cells located at different positions within the same islet (Fig. 3E). Single-cell recordings showed slow and regular plateau depolarizations (Fig. 3E), also known as slow Ca²⁺ waves (18,37). Up to four cells within the same islet were investigated: regardless of their position, intracellular plateau depolarizations were always synchronized with SPs captured via the MEA-electrode located at the bottom of the islet (Fig. 3E). Glucose responses were observed in the majority of cells (63.6%), and 85.7% of them were synchronized with extracellular SPs (Fig. 3F). Hence, multicellular SPs represent summations of synchronized intracellular slow plateau depolarizations of β-cells. These data also indicate that electrical coupling concerns almost the entire islet during the second phase.

We next addressed the relationship between single-cell APs and multicellular SPs and their respective contribution to the biphasic encoding. Low signal-to-noise ratio of metal electrodes at high frequencies prevents the analysis of APs (29). Therefore, we switched to Pedot-MEAs with electroactive polymer-covered electrodes (Fig. 4A), which optimizes AP detection (29) (Fig. 4B). The duration of extracellular APs was ∼100 ms (Fig. 4C), as expected (38). Islets were stimulated by glucose within the narrow physiological range (5.5–8.2 mmol/L), different from supraphysiological levels used in our previous studies (29–32). A measurement of 5.5 mmol/L glucose appeared to be the threshold concentration since a small biphasic response was observed for SPs but not for APs, whereas 6 and 8.2 mmol/L glucose triggered strong biphasic activities for both SPs and APs (Fig. 4D) in ∼90% of islets (Supplementary Fig. 3A) after a short delay (Supplementary Fig. 3B). At 5.5 mmol/L glucose, 45% of islets responded (Supplementary Fig. 3A) with a longer delay (Supplementary Fig. 3B). At this threshold concentration, the presence of SPs, but of small amplitudes and with few APs (Fig. 4D and E), is in line with intracellular recordings in intact islets (39) and may be explained by some β-cells with some K_ATP channel activity at 5.5 mmol/L glucose sufficient to restrain signal propagation.

We also observed further differences in glucose concentration dependency (Fig. 4E). SP amplitudes increased during the first and second phase between 5.5 and 6 mmol/L glucose but remained stable upon a further increase in glucose. In contrast, frequencies of SPs and APs continued to augment until 8.2 mmol/L glucose mainly in the second phase. As maximal SP amplitude represents the size of the functional cell clusters, clusters increase significantly between the two phases for all tested concentrations (Supplementary Fig. 3C). Collectively, these data suggest that increasing glucose from 3 to 8.2 mmol/L recruits more islets and generate more active cell clusters (SP frequency), whereas maximal cluster size is reached already at 6 mmol/L glucose (SP amplitude).

To understand the relative contribution of multicellular and unicellular signals during the transition between first and second phase, SP and AP kinetics were normalized. SP and AP kinetics were similar for the first phase at 5.5 and 6 mmol/L glucose (Fig. 4F, left and middle). Major differences were evident in the second phase at these two concentrations: the small second phase at 5.5 mmol/L glucose involved mainly SPs. Although biphasic patterns of APs started to appear at 6 mmol/L, SPs were still more pronounced especially at the beginning of the second phase as evidenced by differences in the area under the curve (∆AUC) of the respective signals. Moreover, even at 8.2 mmol/L, SPs were far more prominent than APs during the nadir and beginning of the second phase (Fig. 4F, right), a period when insulin secretion persists, albeit at a lower level (Fig. 2). The subsequent development of SPs during the second phase indicates that β-cells synchronize, probably once they are in a metabolic steady state with K_ATP channels blocked to a similar extent. Hence, SP dynamics accurately mirror insulin secretion patterns as compared with APs, confirming that they represent the master electrical signal encoding biphasic secretion.

To ascertain that not only signals from the outer layer of islet cells contribute, we also performed experiments in two-dimensional (2D) monolayers of islet cells (Supplementary Fig. 4). Biphasic behavior was observed in this configuration, but SPs were reduced in the second phase, in line with a 2D coupling in monolayers versus a three-dimensional connectivity in islets (Supplementary Fig. 4A and B). Furthermore, SPs were of higher amplitudes in large monolayers than in medium and small ones, which confirms that SP amplitudes reflect the size of functional β-cell clusters (Supplementary Fig. 4C and D). We would also like to stress that our recordings were performed with physiological glucose and extracellular Ca²⁺ concentrations. Indeed, supraphysiological Ca²⁺ levels exceeding twice the normal concentrations are often used (5,18,37), but create artifactual coupling patterns altering SP and AP dynamics (Supplementary Fig. 5).

Intestinal incretin hormones such as GLP-1 are major modulators of insulin secretion. Effects of postprandial levels of GLP-1 on the two phases of islets activation have never been assessed. Moreover, GLP-1 has rarely been used at physiological picomolar concentrations (32,40), and different pathways may be activated at pharmacological nanomolar concentrations (40). We therefore stimulated the same mouse islets with different glucose concentrations in the absence or presence of 50 pmol/L GLP-1 during both phases. At 5.5 mmol/L glucose, only very weak responses were observed (Fig. 5A) in few islets (Fig. 5C, top), but in the presence of GLP-1, a far greater number of islets became glucose responsive (Fig. 5C, top), and the hormone considerably amplified the second phase (Fig. 5A and D). To confirm that the effect is restricted to the second phase, we examined the effect of picomolar GLP-1 at 8.2 mmol/L glucose. Islets generated two electrical phases for SPs and APs (Fig. 5B), and GLP-1 did not recruit more active islets at this glucose concentration (Fig. 5C, bottom). Again, GLP-1 specifically increased only the second phase and concerned only multicellular coupling signal (SPs) without affecting single-cell activities (APs) (Fig. 4B and E). Note that the activity of islets exhibited oscillations after ∼40 min in G8.2 that disappeared upon GLP-1 (Fig. 5B and Supplementary Fig. 6), confirming a previous observation that the incretin triggers continuous electrical activity in β-cells (40). Moreover, fitting of SP frequencies showed that picomolar GLP-1 accelerated the appearance of the second phase (Supplementary Fig. 6). Thus, postprandial levels of GLP-1 act only on the second phase by enhancing β-cell cluster activity and coupling. GLP-1 action was also examined in human islets. An increase in glucose provoked SPs with a biphasic profile as in mice, which was less pronounced for APs (Fig. 6A). The data confirmed the specific action of GLP-1 in human islets on the second phase of SPs, but not on APs (Fig. 6B).

Aging and type 2 diabetes impair not only the overall quantity of insulin secreted, but also the kinetics (10,13,14). The comparison of SP kinetics between young adult and middle-aged mice revealed that both electrical phases were altered by aging (Fig. 7A): the reactivity of clusters (SP frequencies) was affected without changes in the extent of coupling (SP amplitudes) (Fig. 7B).

Glucotoxicity recapitulates parts of the diabetic state (41). Mouse islets were exposed to a glucotoxic medium (20 mmol/L glucose for 64 h). In these conditions, islets exhibited increased basal SP activities at low glucose (Fig. 7C and D) in line with the increase in basal secretion in glucotoxicity (41). Upon glucose stimulation, the first phase (i.e., high SP frequencies and low SP amplitudes) was considerably altered (Fig. 7D), with a second phase mode starting very early (Fig. 7C). Alterations of biphasic activity were confirmed in human islets exposed to glucotoxicity, and these effects were partially reversible, mainly in terms of coupling signals, as indicated by SPs, but not in regard to single-cell activity (APs) (Fig. 8A and B).

Our data provide a new model for the origin of biphasic islet activation based on analysis of single-cell and of micro-organ electrical activity with high spatiotemporal resolution. Our protocols mimicked relevant physiological characteristics in terms of time spans, concentrations of glucose, GLP-1 (42), as well as Ca²⁺ levels, as the supraphysiological concentrations often used of this cation (5,18,37) considerably distort islet activity.

Our data indicate that progressive multicellular organization establishes the physiological biphasic pattern in both mouse and human islets. Upon glucose stimulation, the first phase originates from a multitude of small β-cell clusters, highly active but poorly coordinated, whereas during the subsequent second phase, clusters enlarge and contain less active but highly synchronized β-cells (Supplementary Fig. 7A and B), in accordance with previous observations using Ca²⁺ imaging in pancreas slices (27). Parallel monitoring of SPs and insulin secretion shows that the overall activity of β-cell clusters in terms of frequency contributes far more than the extent of coupling (given by the amplitude) to biphasic secretion, while the combination of both frequency and coupling is most closely correlated with the biphasic insulin pattern.

The effects of postprandial levels of GLP-1 on both phases were also investigated in this study for the first time. Physiological levels of incretin promote only the second phase by enhancing multicellular signals but not single-cell activities (Supplementary Fig. 7A and B). GLP-1 increases the size of clusters (SP amplitude), their level of activity (SP frequency), and the time spent in active periods. Moreover, GLP-1 also accelerates the installation of the second phase. As the second phase coupling mode constitutes an economic mode, it is certainly more favorable for prolonged activity during the digestion. Its further prominence in the presence of GLP-1 may contribute to protective effects of the incretin on β-cells (43) and certainly underlies its specific secretory effects in vivo on the second phase in humans (44,45). In contrast, aging reduces the reactivity, but not the size, of β-cell clusters, similar to glucotoxicity, for which, in addition, an increased basal activity is observed (Supplementary Fig. 7A and C).

Biphasic hormone secretion has generally been explained by distinct granule pools (19–21). Differential Ca²⁺ sensitivities and kinetics have been observed by intracellular electrophysiology (46), which records, however, only a fraction of the phases. Interestingly, imaging vesicle movements does not unambiguously provide support for distinct vesicle pools as a base for biphasic hormone release (19,47). In addition, biphasic β-cell activity requires multicellular processes (22,23). Multicellular SPs drive the transition between phases, and their profiles clearly mirror the biphasic and monophasic insulin secretions evoked by glucose and leucine, respectively (34). Since electrical SPs occur upstream of exocytosis, our data support the view that SPs propagating across islet β-cells constitute the master regulator of biphasic organization. The other biphasic mechanisms observed downstream at the cytoskeleton and insulin granule levels (5,19–21) may represent a precise adaptation contributing to the amplification of biphasic secretion kinetics.

This dynamic organization appears well adapted to physiological and metabolic requirements. Indeed, the high but poorly synchronized activity of the first phase (Supplementary Fig. 7A and B) provides a rapid and prominent homeostatic response but is rather energy consuming and potentially toxic due to the cytosolic accumulation of Ca²⁺. In the second phase, an increase in coupling concomitant with a reduction of the overall activity prevents such excesses (Supplementary Fig. 7A and B) and may constitute a more economical long-term activity.

The considerable reduction of the first electrical phase upon aging and glucotoxicity (Supplementary Fig. 7A and B) is in line with the clinical data (10–14). Westacott et al. (48) reported that aging alters coupling in mouse and human islets, but the impact on each phase has not been addressed. In the present study, the reduction in SP frequency in the first phase suggests that the basic organizational mode in clusters is not changed per se during this period, but the overall activity of clusters is decreased. Moreover, increase in basal activity blunted the net increment in second-phase activity contributing to the well-known phenomenon of glucose insensitivity (49).

The unbiased long-term approach used in this study provides a new model of islet activation and its derangements. The methodology may be of considerable value to evaluate disease models and maturation; for example, in the setting of normal or patient-obtained stem cell–derived surrogate islets. Finally, better understanding of islet endogenous algorithms as presented in this study may also improve development of new commands driving insulin pumps for the therapy of diabetes (50).

Acknowledgments. The authors thank the colleagues at the University of Bordeaux Animal Facility for help.

Funding. This study was supported by the following grants: European Regional Development Fund Diaglyc (to J.L. and S.R.), Agency Nationale de la Recherche ANR-18-CE17-0005 DIABLO (to J.L. and S.R.), Ministère de l’Education Nationale, de l’Enseignement Superieur et de la Recherche Excellence PhD Scholarship (to M.R. and M.J.), and Université de Bordeaux PEPS Idex/CNRS (to M.R.).

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

Author Contributions. M.J., J.L., and M.R. conceived the study. M.J., E.B., A.P., E.P., J.G., S.O., D.C., and M.R. performed experiments. M.J., E.B., A.P., J.L., and M.R. analyzed data. M.J., J.L., and M.R. wrote the manuscript. F.L. and D.B. provided human islets. B.C., S.R., J.L., and M.R. procured funding. M.R. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented as abstracts and oral communications at the annual meeting of the European Association for the Study of Diabetes, Berlin, Germany, 1–5 October, 2018, Congress of the Société Francophone du Diabète, Nantes, France, 20–23 March 2018, and the Keystone Symposia, Islet Biology: From Gene to Cell to Micro-Organ, Santa Fe, NM, 27–31 January 2020.