Dynamin deficiency causes insulin secretion failure and hyperglycemia

Significance

Insulin secretion regulates broad cellular function and glucose homeostasis, and its disruption causes diabetes. Insulin secretion requires granule fusion with the plasma membrane to allow insulin release (called exocytosis). Exocytosis is coupled with robust endocytosis, a process that retrieves materials from the cell surface. There exists an ongoing debate around the contributions of coupled endocytosis to insulin secretion. Here, we investigate this issue in adult mouse β cells by deleting all three mammalian dynamin genes. Interestingly, dynamin triple-knockout β cells still contain abundant insulin granules but cannot release insulin effectively, causing the disrupted glucose homeostasis in vivo. These findings demonstrate the pivotal role of dynamin in insulin secretion by coordinating the “in” and “out” of membrane trafficking in endocrine cells.

Abstract

Pancreatic β cells operate with a high rate of membrane recycling for insulin secretion, yet endocytosis in these cells is not fully understood. We investigate this process in mature mouse β cells by genetically deleting dynamin GTPase, the membrane fission machinery essential for clathrin-mediated endocytosis. Unexpectedly, the mice lacking all three dynamin genes (DNM1, DNM2, DNM3) in their β cells are viable, and their β cells still contain numerous insulin granules. Endocytosis in these β cells is severely impaired, resulting in abnormal endocytic intermediates on the plasma membrane. Although insulin granules are abundant, their release upon glucose stimulation is blunted in both the first and second phases, leading to hyperglycemia and glucose intolerance in mice. Dynamin triple deletion impairs insulin granule exocytosis and decreases intracellular Ca²⁺ responses and granule docking. The docking defect is correlated with reduced expression of Munc13-1 and RIM1 and reorganization of cortical F-actin in β cells. Collectively, these findings uncover the role of dynamin in dense-core vesicle endocytosis and secretory capacity. Insulin secretion deficiency in the absence of dynamin-mediated endocytosis highlights the risk of impaired membrane trafficking in endocrine failure and diabetes pathogenesis.

Pancreatic β cells are the sole source of endogenous insulin, which dictates glucose homeostasis and diabetes progression. Each β cell contains ∼9,000 to 13,000 insulin granules (1, 2), corresponding to ∼10% to 20% of the cell volume and ∼32 to 46 times the cell surface area. Every meal stimulates insulin granule exocytosis, and resultant cell surface expansion is counterbalanced by endocytosis. Although insulin exocytosis has been extensively studied (3–5), endocytosis in β cells remains poorly understood. Importantly, new progress further indicates the potential role of endocytosis dysfunction in diabetes risk (6, 7).

During glucose-stimulated insulin secretion (GSIS), β cells produce more ATP and depolarize the plasma membrane (PM), and the resultant Ca²⁺ influx triggers SNARE-mediated insulin exocytosis (5, 8). These exocytosis events create a biphasic GSIS pattern. The first phase is transient and involves exocytosis of granules from the readily releasable pool (RRP) (9), whose size is regulated by glucose metabolism (10, 11) and vesicle priming protein Munc13 (12, 13), CAPS (14), and RIM (15, 16). The second phase lasts longer and relies on the mobilization of granules from the reserve pool in the cytosol and subsequent release with or without apparent docking (10, 17, 18). Glucose further amplifies insulin secretion independent of K(ATP) channels (19, 20). Thus, biphasic GSIS is regulated by multiple factors, including insulin granule pools, glucose amplification (11, 19), and actin remodeling (21, 22). Disruption of phasic GSIS is a hallmark of type 2 diabetes (T2D) (22).

Following exocytosis, granule components are recycled back to endosomes and Golgi for degradation or reuse. Although β cell endocytosis was first observed in the early 1970s (23), its molecular basis and physiological function remain poorly defined. Limited studies on this topic show the regulatory role of Ca²⁺ (24, 25), IP₆ (26), and G proteins (27). As a unique form of exo–endocytosis coupling, “kiss-and-run” (28–30) or “cavi-recapture” (31, 32) has been reported in other cell types, but its presence in β cells remains a matter of debate (33–37). Interestingly, diabetes-associated transcription factor Sox4 has been proposed to reduce GSIS by increasing kiss-and-run events (38). The diabetes-risk locus at StAR-related lipid transfer protein 10 (STARD10) is recently associated with FCHSD2 (6), a gene encoding the F-BAR protein required in clathrin-mediated endocytosis (39), and its knockdown reduced GSIS (6). Conversely, perturbation of endocytosis by deleting dynamin-2 in vivo has revealed a causal role of endocytosis in GSIS and glucose intolerance (7). Together with others (40, 41), these findings imply an emerging role of endocytic trafficking in diabetes pathogenesis.

Dynamin is a large GTPase involved in vesicular trafficking, particularly clathrin-mediated endocytosis (42–44). Mammals have three dynamin genes (DNM1, DNM2, and DNM3) with a tissue-specific expression pattern (45). Dynamin-1 and dynamin-3 are highly expressed in the nervous system, and they regulate synaptic vesicle recycling and neurotransmission (46–54). Dynamin-2 is ubiquitously expressed in many tissues, particularly in nonexcitable cells. These distinct expression patterns imply the isoform-specific function of dynamins in different cell types (55). In β cells, dynamin-1 is undetectable and dynamin-2 and -3 are abundant (7). Dynamin-2 selectively regulates the second phase of GSIS, while the dynamin-3 function is yet to be established. It has been reported in other systems that fusion pore reclosing (31, 56, 57) requires dynamin (32, 33, 56, 58, 59).

To fully understand dynamin family GTPases in β cells, here, we have capitalized on triple dynamin-floxed mice (60) and acutely deleted three dynamin genes selectively in β cells in vivo. Surprisingly, β cells after dynamin triple deletion are still viable and produce abundant insulin granules. Nonetheless, dynamin deletion disrupts glycemia, impairs glucose tolerance, and significantly reduces GSIS. At the cellular level, both endocytosis and insulin exocytosis are impaired. This study provides genetic evidence on the role of dynamin-dependent membrane recycling in β cells and its implication in diabetes pathogenesis.

Results

Acute Triple Dynamin Deletion in Adult β Cells Impairs Blood Glucose Homeostasis.

We have generated β cell-specific and tamoxifen-inducible dynamin triple knockout (TKO) mice by crossing dynamin-floxed mice (DNM1^f/f, DNM2^f/f, DNM3^f/f) (60) with Tg(Ins1-Cre/ERT)1Lphi (MIP1-CreERT) mice (61, 62). TKO mice (DNM1^f/f, DNM2^f/f, DNM3^f/f, MIP1-CreERT) were visually indistinguishable from their littermate controls. We chose MIP1-CreERT mice because they express Cre only in β cells (61) and have no spontaneous expression before tamoxifen administration. The latter feature is crucial, as conventional dynamin-2 knockout (KO) is embryonically lethal (63). Although MIP1-CreERT mice contain human growth hormone (hGH) (64) similar to other β cell Cre lines (65), they show intact islet architecture, GSIS, and glucose tolerance (62, 64). We have confirmed the unchanged glucose tolerance in different groups of control mice that either lack or carry MIP1-CreERT (see Fig. 2A). Thus, we use TKO littermates (DNM1^f/f, DNM2^f/f, DNM3^f/f treated with tamoxifen) as control mice hereafter.

Time-controlled, β cell-specific dynamin KO is achieved by tamoxifen injection in adult mice (see Methods), and dynamin depletion is confirmed by immunoblots and immunofluorescence imaging (Fig. 1 A and B). Since dynamin-1 protein is undetectable in mouse β cells (7), we focus on dynamin-2 and dynamin-3. Western blots revealed a substantial decrease of dynamin-2 in TKO islet lysates (Fig. 1A) (n = 6 to 7 independent tests). The remaining faint signal is expected and presumably attributed to a small number of non–β cells in islets such as α cells, δ cells, pancreatic polypeptide cells, and vasculatures. The effective depletion of dynamin-2 and dynamin-3 is demonstrated by immunofluorescence detection in TKO islet sections in situ (Fig. 1B) and further confirmed in isolated single β cell cultures (SI Appendix, Fig 1). These data suggest that this approach effectively deletes dynamin in β cells in vivo.

Fig. 1.

Acute β cell–specific triple dynamin deletion in adult mice led to hyperglycemia. (A) Western blots of purified islets from control and dynamin TKO mice with anti–dynamin-2 (anti-dyn2; rabbit polyclonal) and actin. The right panel shows the quantification of Western blots (as dyn2/actin ratio). P = 0.001, Student’s t test (two-tail); mean ± SEM; n = 7 and 6 blots for control and TKO, respectively; each blot contains islets from 1 to 2 mice. (B) Dynamin immunofluorescence of pancreas islet sections. (Scale bar, 100 μm.) Three mice per genotype. Anti-dyn2 is rabbit polyclonal, and anti-dyn3 is 5H5 mouse monoclonal (costain islet vasculatures and dyn3 in β cells). (C) Increased blood glucose levels in free-feeding TKO mice. Glucose was measured 40 d after the last tamoxifen injection (n = 18 control and 21 TKO mice, P = 0.003, t test). (D) Body weight of control and TKO mice (n = 18 control and 21 TKO mice, P = 0.78, t test). (E) Glucose increased 10 d after tamoxifen injection (n = 9 to 14 mice per group as indicated by individual points. Data at the day-40 group were from C). *P < 0.05, **P < 0.01, ***P < 0.005.

TKO mice exhibited higher blood glucose levels than controls (Fig. 1C) (116.7 ± 4.8 mg/dL; n = 18 mice for control; 138.3 ± 4.9 mg/dL; n = 21 mice for TKO; P = 0.011; Student’s t test), without changing body weight (Fig. 1D). Hyperglycemia started 10 d after the last tamoxifen injection (Fig. 1E), indicating time dependence of dynamin deletion in vivo. It is of interest to note that overnight-fasting glycemia was intact in TKO mice (Fig. 2A). Further, TKO impaired glucose tolerance (Fig. 2A) without affecting insulin tolerance (Fig. 2B). Under the metabolic stress of a high-fat diet (for 3 mo), TKO mice developed more severe glucose intolerance than those fed with a normal diet (Fig. 2 C, Right, P < 0.01), but the changes in insulin tolerance and body weight increase were comparable to controls (Fig. 2 D and E). These data demonstrate that dynamin deletion selectively impairs β cell function rather than affecting peripheral tissues.

Fig. 2.

Dynamin TKO disrupted mouse glucose homeostasis and β cell mass. (A) Glucose tolerance tests (GTTs) and their area under curve (AUC) analysis (n = 6 to 8 per group, one-way ANOVA with post hoc Tukey honest significant difference test, *P < 0.05, **P < 0.01, ***P < 0.005). In the legend, Flox = DNM1^f/f 2^f/f 3^f/f, V = vehicle (corn oil), and T = tamoxifen injection. (B) Insulin tolerance tests (ITTs) were intact in TKO mice (n = 6 control and 8 TKO mice, t test, P > 0.05). (C) GTTs and AUC analysis in mice fed with high-fat (HF) diet (HFD) (P < 0.005) (n = 6 control and 11 TKO mice. Age-matched mice were fed with HFD for 3 mo). (D and E) ITTs in HFD mice (D) and body weight increase (E) (n = 6 control and 11 TKO mice, t test). (F) Islet cell mass in control and TKO mice (n = 7 mice per group, P < 0.05 for β cells, two-tailed t test). Each point represents one mouse. (G) Representative images of α cell redistribution in a control islet and a TKO islet. Note the α cells in the core of the TKO islet. (Scale bar, 50 μm.)

To discern potential β cell loss, we quantified β cell mass in TKO mice using the large-scale imaging of the whole pancreas (66, 67), which avoids the bias of regional heterogeneity of islet density. We found a slight reduction of β cell mass despite the large interpancreas variability (n = 7 mice per group, P > 0.05, t test) (Fig. 2E). α and δ cell mass remain unchanged. Islet sizes were highly variable (ranging from tens to hundreds of micrometers in diameter) (SI Appendix, Fig. 3), and average islet numbers and sizes in TKO were decreased. These data suggest that β cell loss partially contributes to the in vivo phenotypes. Besides, we observed that α cells were located at the center of some TKO islets (Fig. 2G). A similar change has been reported in dynamin-2 KO mice (7) and some rodent models of diabetes (68, 69), presumably indicating altered intra-islet signaling or metabolic stress.

Triple Dynamin KO Impairs Both Phases of GSIS.

To evaluate dynamin function in GSIS, we performed islet perfusion to measure insulin secretion. In control islets, 20 mM glucose evoked biphasic GSIS (Fig. 3A, black), similar to our previous work (7). The first phase peaked at ∼5 min after stimulation and declined within another 5 to 10 min, and the second phase developed slowly but lasted longer. In TKO islets, GSIS decreased significantly in both phases (P < 0.05, 30 islets/perifusion/mouse, n = 6 mice per genotype, two-tailed t test) (Fig. 3A). Insulin secretion after 30 mM KCl was also decreased (Fig. 3B, n = 6 mice per group, P < 0.01, two-tailed t test). Therefore, dynamin deletion impairs both the first and second phases of GSIS, consistent with hyperglycemia in TKO mice.

Fig. 3.

Dynamin TKO impaired both phases of GSIS and glucose-induced Ca²⁺ signaling. (A) Impaired GSIS in the first and second phases from TKO islets (30 islets/group/test, 6 independent experiments for each genotype). The right panel shows the area under curve (AUC) analysis of the first phase (10 to 20 min, P < 0.05, two-tailed t test) and the second phase (20 to 40 min, P < 0.01, two-tailed t test, six independent assays per group). (B) Reduction of depolarization-induced insulin secretion. Islets were stimulated with 30 mM KCl for 10 min (n = 6 independent experiments per group). The right panel shows its AUC analysis (P < 0.01, two-tailed t test). (C) Representative dynamics of β cell [Ca²⁺]_i in control (black) and TKO (red) islets. (D) Quantification of [Ca²⁺]_i amplitudes in control and TKO islets (n = 24 islets in control and 31 islets in TKO, three mice per genotype; P < 0.005, two-tailed t test). Note the smaller peak [Ca²⁺]_i induced by 12.8 mM glucose in TKO β cells. (E) NADPH fluorescence in islets at rest and after 20 mM glucose stimulation (n = 15 and 14 islets in control and TKO group; P < 0.005, three mice per group). *P < 0.05, **P < 0.01, ***P < 0.005.

TKO Disrupts Glucose-Induced Ca²⁺ Signaling in β Cells.

How does dynamin regulate GSIS? One possibility is the depletion of insulin granules due to impaired dynamin-mediated endocytosis. However, TKO β cells still contain numerous granules (see Fig. 6 below), implying that defects at the secretory pathway’s late steps impair insulin secretion.

Fig. 6.

Dynamin is required for maintaining insulin granule density and granule docking in β cells. (A) β cell EM graphs from control and TKO islets. Note the abundant granules in the cytosol and reduction of docked granules (asterisks) on the PM in TKO β cells. Endocytic intermediates (arrows) and some large vacuoles are also visible in the right panel. (Scale bar, 1 μm.) (B–D) Quantification of insulin granule size (B), density (C), and spatial distribution away from the PM (D) (***P < 0.005, 200 nm binning). Data were from n = 9 to 15 EM sections randomly collected in islets from each group; two-tailed t test. Three mice per genotype were tested for EM assays. Granule distribution is defined by the distance between the PM and the closest point of each granule. (E) Changes in docked granule number in control and TKO β cells at rest (5.5 mM glucose) and after glucose stimulation (20 mM for 20 min) (n = 9 to 15 EM sections, one-way ANOVA, ***P < 0.005). Granule docking was defined by the presence of direct granule–PM contact in EM images. (F) Granule distribution away from the PM after glucose stimulation (20 mM, 20 min). *P < 0.05, **P < 0.01.

Because intracellular Ca²⁺ signal triggers exocytosis and effectively modulates vesicle fusion in a superlinear fashion (70–72), we examined if this step was affected in TKO β cells. Fig. 3 shows the Ca²⁺ signal elicited by glucose and K⁺ under ratiometric Ca²⁺ imaging. While basal Ca²⁺ levels were similar between control and TKO islets, peak Ca²⁺ amplitudes induced by 12.8 mM glucose were significantly smaller in TKO (P < 0.001) (Fig. 3D), suggesting altered stimulation-secretion coupling during glucose stimulation. Using NADPH autofluorescence as a readout of glucose metabolism (7), we found that glucose increased islet NADPH levels (P < 0.005) in both control and TKO islets with similar amplitudes (P > 0.05) (Fig. 3E). These data suggest that TKO impairs glucose-induced Ca²⁺ signaling downstream glucose uptake and metabolism (e.g., K[ATP] channels, Ca²⁺ influx, or intracellular Ca²⁺ stores). We further examined the synchronization of Ca²⁺ oscillations in islets, as its disruption by either lipotoxicity or silencing islet “hub-cells” can impair GSIS (73, 74). In both control and TKO groups, Ca²⁺ activity from β cells at spatially separated islet regions was synchronized, and we observed no apparent synchronization defect in TKO islets (SI Appendix, Fig. 5).

Dynamin-Dependent and Dynamin-Independent Endocytosis in β Cells.

We are keen to see what happens to β cell endocytosis without all three dynamin genes since dynamin is thought to be essential for endocytosis in many cells, including endocrine cells. First, we measured endocytosis using real-time membrane capacitance (C_m) recordings (47), as shown in Fig. 4A. The C_m jump and subsequent decay reflect exocytosis and endocytosis, respectively. Consistent with the prior work (7), C_m decay kinetics was heterogeneous among individual cells. The average C_m trace in control cells decayed rapidly and was best fitted by double-exponential (black, τ₁ = 2.4 s, τ₂ = 8.6 s, n = 6 cells). In contrast, C_m in TKO β cells decayed much slower (red, τ₁ = 4.5 s, τ₂ = 51 s, n = 10 cells), suggesting impaired endocytosis. Notably, C_m decay was not abolished by TKO, and it still recovered ∼20% of controls on average (Fig. 4B), suggesting the contribution of dynamin-independent endocytosis. This finding may explain why TKO β cells were still viable and why TKO mice can survive.

Fig. 4.

Triple dynamin deletion severely impaired β cell endocytosis. (A) Disrupted endocytosis in β cells as measured by C_m decay. C_m traces were averaged from n = 6 control and n = 10 TKO β cells and presented as mean ± SEM. Data were from five mice per group. The top panel shows the depolarization scheme. (B) Normalization of C_m traces. Note dynamin-dependent and dynamin-independent endocytosis in TKO β cells. Individual C_m traces were normalized to their peak, averaged, and fitted with a double exponential function (blue lines). (C) C_m dynamics during the train stimulation in control and TKO (n = 22 and 15 cells for control and TKO group). The stimulus contains a train of 10 pulses (500 ms, 0 mV) at 1 Hz. (D) Endocytosis rate following each pulse during the stimulation shown in C. (E) Impaired transferrin uptake in TKO β cells. Transferrin fluorescence intensity was quantified (right) (n = 6 to 10 cells per group per cultures and total of three independent cultures. P = 0.003, two-tailed t test). (F) Accumulated transferrin at the surface of TKO β cells (without acid washing). The arrow indicates cytosolic transferrin, and the arrowhead points to enhanced transferrin (red) and clathrin (blue) signal on the PM. The right panel shows transferrin intensity at three-dimensional plot. ***P < 0.005.

Dynamin-1 at synapses is reported to function selectively during stimulation (46). We tested this in TKO β cells using a burst of electrical pulses (10 pulses at 1 Hz, 0.5 s 0 mV for each pulse) during capacitance recordings (47). As reported previously in β cells (7), two major types of Cm dynamics were observed (SI Appendix, Fig. 4A): 1) C_m elevation without any sign of C_m decrease (12 out of 22 cells) and 2) C_m elevation mixed with C_m decrease between pulses. On average, the control C_m trace increased initially but decreased after the fourth pulse. In contrast, the TKO C_m trace exhibits little endocytosis (Fig. 4C, n = 22 control cells, n = 15 TKO). Quantitative analysis (SI Appendix, Fig. 4 B and C) revealed a higher endocytosis rate in control than TKO β cells. This result suggests that TKO β cells cannot keep up with the pace of exocytosis load because of impaired endocytosis.

Second, we examined clathrin-mediated endocytosis. Fig. 4E shows that TKO significantly decreases transferrin uptake in β cells. In this assay, internalized transferrin was measured after removing surface transferrin by acid wash. If we skipped the acid-wash step, transferrin was primarily accumulated on the PM of TKO β cells and colocalized with clathrin (Fig. 4F). These data suggest the impairment of clathrin-mediated endocytosis in TKO β cells.

Third, electron microscopy (EM) revealed severely distorted endocytosis intermediates adjacent to the PM in TKO but not control β cells. Fig. 5A shows a gallery of clathrin-coated pits (CCPs) and abnormal endocytosis intermediates in the TKO group. They display varied morphology and size, and many of them have elongated fission necks or long membrane tubules connected to extracellular space. Abnormal intermediates with more complex structures were occasionally captured within a single EM section (Fig. 5A, TKO 5 through 9) but were absent in controls. The CCP number of TKO is significantly higher than control and dynamin-2 KO (Fig. 5B). The worsening endocytic disruption suggests a partially overlapping role between dynamin-2 and -3 in β cell endocytosis.

Fig. 5.

Accumulation of abnormal endocytic intermediates in TKO β cells. (A) Gallery of EM images on endocytic intermediates in control and TKO β cells. Note CCPs in TKO formed more complex structures, some of which had a larger size (4, 5, and 8), a long tubule (6), or a string of CCPs (5, 7, and 9). (B) Quantification for CCP number (mean ± SEM) in β cells at rest and after 20 mM glucose stimulation for 20 min (20G). Cell numbers were indicated in each bar (one-way ANOVA, *P < 0.05; **P < 0.001). Six to eight islets were examined for each condition; data were pooled from two mice for each group. (C) A single section of EM tomogram in a TKO β cell showing the accumulation of CCPs and disrupted endocytosis intermediates (arrows) adjacent to the PM. The islets were stimulated with 20 mM glucose stimulation for 20 min. (D) Three-dimensional (3D) segmentation for the EM tomogram in C. Note the abundant CCPs (red) and the elongated, narrow tubules (light green) that connect CCPs to the PM (dark transparent green) directly or through a membrane expansion. (E–H) A gallery of abnormal endocytic intermediates from the 3D structures in D. (E) Three CCPs connected on the same PM invagination. (F) A single CCP with a long tubule to the PM. (G) Two CCPs formed on a long tubular PM structure. (H) A big CCP with a thin neck on the PM. Similar structures were absent in control cells. (Scale bars: 200 nm in A, C, and D; 50 nm in E–H.)

Lastly, EM tomography further uncovered three-dimensional ultrastructures of endocytic intermediates in TKO β cells after stimulation (20 mM glucose, 20 min) (Fig. 5 C–H). Accumulated CCPs were connected with the PM through elongated fission necks or narrow tubules (20 to 80 nm) (also see refs. 7, 63). The tubules can be long, expanded, or branched. These ultrastructures are consistent with the accumulation of transferrin on TKO PM (Fig. 5F). They are similar to those in dynamin-2 KO β cells but less complex than those at neuronal synapses (46, 50, 75), suggesting the differences of endocytosis between dense-core vesicles and synaptic vesicles.

The functional and structural assays mentioned demonstrate severe endocytosis defects in TKO β cells. While these data speak for the crucial role of dynamin family GTPases in β cell endocytosis, dynamin-independent endocytosis, albeit less efficient, mitigates the imbalanced endocytic trafficking in TKO β cells.

TKO Impairs Insulin Granule Docking and RRP Size.

We examined ultrastructure changes along with the granule secretory pathway (Fig. 6). Islet β cells under EM were identified by characteristic halo-patterned insulin granules (7). Notably, TKO β cells still contained abundant insulin granules (Fig. 6A) contrasting with the vesicle depletion observed at mouse synapses lacking DNM1 and DNM3 (50, 51) and at Drosophila synapses of Shibire (76), a temperature-sensitive mutation of the only dynamin homolog in flies. Again, TKO cells contained more frequent endocytic membrane invaginations (white arrows). Insulin granule sizes were unchanged (Fig. 6B), but granule density was reduced (Fig. 6C). Spatial analysis of granule distribution revealed ∼50% reduction of granule numbers in the regions within 200 nm distance from the PM (Fig. 6D). The number of docked granules (asterisks in Fig. 6A, defined by direct granule-PM contact) decreased dramatically in resting TKO β cells (P < 0.001, Fig. 6E), and this defect was partially mitigated after glucose stimulation (20 mM for 20 min).

Total internal reflection fluorescence (TIRF) microscopy offers another method to estimate granule distribution near the PM since it selectively images granules within ∼120-nm space beneath the PM. Consistent with the EM data mentioned, TIRF microscopy revealed decreased granule density on TKO β cells’ PM (Fig. 7A). Both methods consistently suggest that TKO impairs insulin granule docking, congruent with reduced GSIS in the first phase (Fig. 3A).

Fig. 7.

Dynamin TKO impaired granule exocytosis and RRP size in β cells. (A) Insulin granules in control and TKO β cells are visualized with an anti-insulin antibody under TIRFM. (B) Reduced granule density in TKO β cells under TIRFM (n = 11 and 17 cells for control and TKO group, P < 0.005, two-tailed t test, repeated in three independent cultures). (C) Impaired exocytosis during a brief pulse (top, 200 ms, 0 mV) in TKO β cells. Exocytosis is measured as the C_m increase (average trace from 10 cells). (D) Reduced RRP size estimated by a step depolarization (1 s, 0 mV) in TKO β cells. (E) Quantification of exocytosis following a pulse of 30 ms, 200 ms, and 1 s depolarization (P < 0.05, two-tailed t test; n = 12 to 14 cells per genotype for 30 ms, n = 10 for 200 ms, and n = 6 to 16 cells for 1-s pulse. Data are pooled from five mice per genotypes). (F and G) Exocytosis induced by a burst of physiologically relevant electric pules (10 × 30 ms at 0 mV at 100-ms interval) and its quantification (n = 23 and 12 cells for control and TKO, P < 0.01 at each point in G, two-tailed t test, five mice per genotype). *P < 0.05, ***P < 0.005.

We next assess the release competence of insulin granules and functional granule pools using C_m recordings. We probe granule release competence with a brief pulse, which only partially releases the immediately releasable pool (IRP) adjacent to Ca²⁺ channels (70, 77–79). A 30-ms pulse induces only a slight C_m elevation in control β cells, and this value is reduced in TKO (Fig. 7E). Similarly, C_m elevation after a 200-ms pulse also decreased significantly in TKO (Fig. 7C). These data suggest a lower release probability of insulin granules in TKO β cells. In response to a longer pulse (1 s, 0 mV step depolarization), which is strong enough to deplete all the release-competent granules, C_m elevation in TKO is smaller than controls, indicating an RRP reduction (Fig. 7 D and E, P < 0.05). These data demonstrated functional impairment of RRP in TKO. Lastly, in response to a train of brief pulses (10 × 30 ms, 0 mV pulses at 100-ms interval) that mimic physiological action potentials in islet β cells (Fig. 7 F and G), TKO β cell displayed an ∼50% reduction of exocytosis. Collectively, these data suggest that dynamin deletion reduces the sizes of IRP and RRP in β cells, and the decline is attributed, at least in part, to disrupted granule docking.

To gain further insight into impaired granule docking and RRP, we turn to molecules known to regulate vesicle docking. First, we examined SNAP-25 and syntaxin-1, as these two SNARE proteins promote vesicle docking (80–82) and their expression is reduced in β cells from diabetic Goto-Kakizaki rats (83, 84) and patients with T2D (85). However, their expression levels did not change in TKO islets (SI Appendix, Fig. 6 A and B). Second, we measured Munc13-1 and RIM1, two proteins initially discovered as priming factors (86–90) and later recognized to regulate vesicle docking (91–93). Our immunodetection data revealed their high expression in wild-type β cells (Fig. 8), consistent with previous works (13, 16, 94, 95). Interestingly, both protein levels were decreased in TKO β cells (Fig. 8 A and B, P < 0.01, three independent tests), suggesting their role in the disrupted granule docking. Third, we examined β cell cortical F-actin, a critical factor in regulating insulin granule mobilization (7, 96) and docking (97). We found a substantial reorganization of cortical F-actin in TKO cells (Fig. 8C). The thin layer of the F-actin network in control β cells was replaced by a denser, thicker layer of actin web in TKO cells. This result is more severe than in single dynamin-2 KO β cells (7), indicating an additional role of dynamin-3 in this process. Together, the reduction of Munc13-1 and RIM1 expression and cortical actin disorganization can contribute to the defects of insulin granule docking and subsequent secretion in TKO β cells.

Fig. 8.

TKO decreased Munc13-1 and RIM1 expression and disrupted cortical F-actin in β cells. (A) Reduced fluorescence signal of anti-Mucn13-1 and anti-RIM1 in TKO β cells revealed by confocal microscopy. (Scale bars, 10 μm.) (B) Quantification of fluorescence intensity in A. n = 157 control and 268 TKO β cells for Munc13-1 assays; n = 92 control and 207 TKO for RIM1 assays; two-tailed t test, P < 0.005; presented as mean ± SEM. Data were from three independent cultures per genotype. (C) Cortical F-actin reorganization in TKO β cells under TIRFM. (Scale bars, 5 μm.) Inserts show the expanded views of cortical F-actin in the box regions. (D) Quantification of F-actin intensity under the PM in a representative culture (n = 9 and 17 β cells for control and TKO groups, two-tailed t test, P < 0.005. Total of four independent cultures per genotype). ***P < 0.005.

Discussion

The mechanism of endocytosis in dense-core secretory cells is obscure. Here, we use a genetic approach to define the role of dynamin-mediated endocytosis in β cells. We found that dynamin-null β cells still contain abundant insulin granules, but they are reluctant to release insulin, leading to hyperglycemia and glucose intolerance in TKO mice. Furthermore, dynamin deletion impairs both clathrin-mediated endocytosis and dense-core secretory pathway, resulting in reduced GSIS in both the first and second phases. These findings provide compelling evidence that triple-dynamin deficiency in β cells can cause endocrine failure in vitro and in vivo.

The genetic study on hormone secretion in the absence of all three dynamin genes has not been conducted before. Interestingly, dynamin-null β cells generated here still contain many insulin granules. Capacitance recordings demonstrate that endocytosis still resumed, albeit much less efficiently, suggesting the presence of dynamin-independent endocytosis. Dynamin has long been regarded as the membrane fission machinery of endocytosis (42, 43, 98), but increasing evidence indicates that dynamin-independent endocytosis also occurs (60, 99–101). Dynamin-like Vps1 in yeast is dispensable in endocytosis (102). Shibire mutants show intact fluid-phase endocytosis (103) and bulk endocytosis (104). Dynamin dependence of endocytosis in mammalian secretory cells is inconclusive so far, partially because of the presence of multiple dynamin genes (43, 45) and certain functional redundancy (105). Deletion of both dynamin-1 and -3, which account for >90% of total dynamin in neurons (46), impairs clathrin-mediated endocytosis (50, 51, 105) but not bulk endocytosis (75), implying the possibility of dynamin-independent endocytosis or compensation of the remaining dynamin-2. Conditional dynamin-2 deletion alone shows little effect at the calyx of Held synapses but exacerbates the defects of dynamin-1,3 double KO in a gene dose-dependent manner (105), and triple dynamin deletion causes neuron death at an early developmental stage (105). Thus, a definite assessment of dynamin function in secretory cells is still missing.

While neurons cannot survive in vivo without dynamins (105), β cells in the TKO mice are viable and contain abundant granules. Thus, these cells offer an unprecedented opportunity to conclusively study dynamin function in secretory cells, which is otherwise unattainable. Further, the data presented here suggest the operation of both dynamin-dependent and dynamin-independent endocytosis. The former ensures optimal β cell function and GSIS, whereas the latter can support TKO β cell survival and less effective insulin release. Thus, while dynamin regulates β cell endocytosis, it is by no means indispensable.

Endocytosis study in β cells dates back to the early 1970s (23), yet there is a striking paucity of research on its molecular mechanism and physiology, particularly in the context of GSIS and glycemic regulation. On the other hand, recent progress implies the potential role of endocytosis in diabetes. Studies on diabetes-associated genes SOX4 (38), FCHSD2 (6), and NHA2, a Na⁺/H+ exchanger associated with diabetes (106), indicate their roles in β cell endocytosis. Direct endocytosis perturbation (7) by the dynamin-2 deletion in mice disrupts insulin secretion and causes glucose intolerance. Pancreatic β cells in rodents express high levels of dynamin-2 and dynamin-3 but not dynamin-1 (7). Dynamin-2 selectively regulates the second GSIS phase, but the role of dynamin-3 is unknown. The current study uses conditional triple dynamin KO to study endocytosis in β cells. The exacerbated phenotype in TKO β cells rather than dynamin-2 KO suggests a redundant role of dynamin-3, consistent with the regular appearance of dynamin-3 single KO mice (50). Despite the growing evidence for the connection between β cell endocytosis and T2D, dynamin variants associated with T2D have not been reported thus far in human genome-wide association studies (https://t2d.hugeamp.org/). It is unclear whether the lack of association is relevant to the embryonic lethality of DNM2 disruption, isoform redundancy, or lower minor allele frequency (MAF) of variants in GWAS analysis. Dynamin mutations have been identified in Charcot-Marie-Tooth peripheral neuropathy (CMT) and autosomal dominant centronuclear myopathy (107), and the copresence of T2D with CMT has been reported in some patients (108–110), raising the question whether dynamin dysfunction is the shared etiology for both diseases in these cases.

Hyperglycemia and glucose intolerance in the TKO mice (Figs. 1 and 2 A–C) speak for dynamin function in glycemic homeostasis. These defects attribute to both β cell dysfunction in GSIS (Figs. 3–8) and the reduction of β cell mass (Fig. 2F). Interestingly, dense core granules are still abundant in TKO β cells, but they cannot be released effectively. The abundance of granules observed here contrasts with largely depleted synaptic vesicles in DNM1 and DNM3 double KO synapses, suggesting the different recycling pathways for dense core vesicles and small synaptic vesicles. TKO β cells moderately decrease their insulin granule density, but this level of reduction appears unlikely to be a rate limit for GSIS. Further morphological and functional studies reveal the disruption of phasic insulin secretion in both the first and second phases. These defects result from the reduced glucose-elicited [Ca²⁺]_i amplitude, RRP size, and granule docking.

In search of molecules behind the disrupted granule docking and RRP, we have screened some docking-associated proteins. We found that SNAP-25 and syntaxin-1 remain intact but Munc13-1 and RIM1 levels are strongly reduced (Fig. 8). The latter findings imply previous unknown signaling between endocytosis and exocytosis that deserves further investigation. In addition, TKO drives a significant cortical F-actin reorganization in β cells, which is consistent with dynamin-actin interactions (111) and the recently identified role of dynamin in regulating actin network architecture, by coordinating the activity between F-actin bundling and Arp2/3-mediated branching (112). Together, these findings contribute, at least in part, to impaired GSIS in TKO β cells.

Earlier literature, which is often based on studies in cell lines using dominant-negative or silencing approaches (33, 41), has reported different effects on insulin release upon reducing dynamin activity. The results may be related to kiss-and-run, a near-synchronous recapture of granules after fusion pore formation and cargo release. This process is opposed to full fusion, which involves the complete collapse of the granule membrane into the PM. Its importance remains unresolved in β cells (33–37). Notably, dynamin is thought to limit fusion pore expansion and cargo release during insulin granule exocytosis (33), but dynamin K44A perturbation reduces secretion (41). GLP-1–evoked cAMP signaling increases insulin secretion but paradoxically shifts full fusion to kiss-and-run (113, 114). Here, we take a genetic approach to remove dynamin and thus dynamin-dependent endocytosis. Multiple lines of data provided here demonstrate a prominent impairment of insulin secretion in the absence of dynamin. While dynamin regulates endocytosis and endocytosis-associated cell signaling (115) to impact insulin secretion, its role in other intracellular processes (43, 116), including fusion pore dynamics, cannot be ruled out and deserves future investigation.

In summary, this work provides compelling genetic evidence on dynamin-mediated regulation in β cell endocytosis, GSIS, and glycemia, implying the risk of disrupted membrane trafficking homeostasis in β cell failure and diabetes.

Methods

Generation of β Cell–Specific, Tamoxifen-Induced Dynamin TKO Mice.

The conditional dynamin KO mice were generated by crossing triple dynamin floxed mice (DNM1^f/f, DNM2^f/f, DNM3^f/f) (60) with Tg(Ins1-Cre/ERT)1Lpi (also named MIP1-CreERT) transgenic mice (62) to obtain the needed genotype (DNM1^f/f, DNM2^f/f, DNM3^f/f, MIP1-CreER). Genotypes were determined by PCRs. The primer sequences were 5′-GCA GGA AGA CAC ACA ACT GAA C -3′ and 5′-CCT GCT AGT GAC CTT TCT TGA G-3′ for DNM2^f/f gene, 5′-CAG TGC CTT CCA AGT TCA ATT CC-3′ and 5′-GAC ATG TTA ACA TAG GCT AAA CC-3′ for DNM3^f/f gene, 5′-TTG TGT ATG TGA GTG CAC CCA TGC-3′ and 5′-CAG CTG GGT ATA ATG AGG CCT CAT C-3′ for DNM1^f/f gene, and 5′-CCT GGC GAT CCC TGA ACA TGT CCT-3′ and 5′-TGG ACT ATA AAG CTG GTG GGC AT-3′ for MIP1-CreERT. Conditional TKO mice before tamoxifen administration showed no overt difference from their littermate controls.

Age-matched adult mice (2 to 6 mo old), both male and female, were used. To achieve β cell–specific dynamin deletion in the conditional TKO mice (DNM1^f/f, DNM2^f/f, DNM3^f/f, MIP1-CreER), tamoxifen was injected intraperitoneally (0.15 g/kg body weight, dissolved in corn oil as the vehicle, once a day for 10 d). Experiments started ∼4 wk after the first tamoxifen injection. TKO mice and their littermate controls (DNM1^f/f, DNM2^f/f, DNM3^f/f) were paired for tests whenever possible. Despite the fact that MIP1-CreER mice express hGH and are resistant to streptozotocin-induced diabetes (64), they have intact islet structures, GSIS, and glucose tolerance (7, 62, 64). Data in Fig. 2A using three different control groups demonstrated comparable glucose tolerance test results regardless of the presence of MIP1-CreER, suggesting the specific role of dynamin deletion. This result is consistent with prior studies from us (7) and others (62, 64). Thus, we used the DNM1^f/f, DNM2^f/f, and DNM3^f/f mice (with tamoxifen) as controls unless specified.

Western Blotting.

Purified islets were homogenized in lysis buffer and followed by Western blotting. Primary antibodies include actin (Millipore, 69100, mouse), GAPDH (Thermo Fisher Scientific, MA5-15738), α-tubulin (Santa Cruz, sc8035), Syntaxin-1 (Santa Cruz, sc-12736), SNAP-25 (Covance, SMI-81R), and dynamin 2 (rabbit polyclonal, from Pietro De Camilli). Horseradish peroxidase-conjugated secondary antibodies were used for chemiluminescent imaging. Each test was repeated at least three times.

Immunofluorescence Staining.

Pancreata were fixed with 4% paraformaldehyde (PFA) + 4% sucrose in 0.12 M sodium phosphate buffer, embedded in the optimal cutting temperature compound (OCT), and cut into 8 µM frozen sections before immunostaining. Cultured cells on coverslips were fixed with 4% PFA. Samples were immunostained by primary antibodies and the fluorescence conjugated secondary antibodies as we previously described (7).

Primary antibodies used in this study included insulin (Abcam ab7842, guinea pig), glucagon (Abcam, ab10988, mouse), clathrin light chain (Millipore, AB9884), SNAP-25 (Covance, SMI-81R), Munc13-1 (SYSY, rabbit polyclonal, 126102), RIM1 (SYSY, 140003), dynamin-1 (rabbit monoclonal, EP801Y, #1851-1 Epitomics), dynamin-2 (rabbit polyclonal antibody from Pietro De Camilli at Yale University) (Fig. 1 A and B), and dynamin-3 (5H5, mouse monoclonal from Pietro De Camilli) (50). The specificity of dynamin antibodies has been previously verified with conventional dynamin KO (7). SI Appendix, Fig.1B used G-4 mouse monoclonal anti-dynamin-2 (sc-166669, Santa Cruz). Alexa Fluor 488 Phalloidin (A12379, 300 unite), Alexa Fluor 647 Phalloidin (A22287, 300 unit), and the secondary antibodies conjugated with different Alexa Fluorophores (AF-405, AF-488, AF-568, AF-594, and AF-647) were from LifeTechnology.

Spinning Disk Confocal and TIRF Microscopy.

Multiple-color fluorescence imaging was performed using a Nikon Ti-E–based confocal, TIRF, and single-molecular localization imaging system (117). The system is equipped with Ti-ND6-PFS Perfect Focus, a motorized nosepiece with multiple objectives (CFI Plan Fluor 4×, NA 0.13 WD, 17.2 mm; CFI Plan APO Lambda 20×, NA 0.75, WD 1.00 mm; Plan APO VC 60× Oil, NA 1.4, WD 0.13 mm and APO 100× Oil, NA 1.49, WD 0.12 mm), a spinning disk (CSU X-1, 10,000 rpm, Yokogawa), Ti-TIRF Motorized Illuminator Unit, Agilent MLC400 High Power Monolithic Laser Combiner SP with 405, 488, 561 and 642 nm lasers, electron-multiplying charge-coupled device (EMCCD, iXon ×3 DU897, Andor) and Neo-sCMOS cameras (Andor) lateral imaging ports of the microscope, 10-position filter wheels (Sutter), and Lumen 200 Illumination System. Spinning disk confocal images were acquired at the left imaging port, and TIRF images were obtained at the right port through the 100× objective (NA = 1.49) under the critical illumination angle.

Imaging analysis and quantification were performed using NIS-Elements AR (Nikon). Images from control and TKO cells for comparison were acquired and displayed using the same settings. The insulin granules under the TIRF field were counted using the spot detection function in NIS-Elements AR.

Glucose Tolerance Tests and Insulin Tolerance Tests.

Blood glucose concentration was determined from mouse tail-vein blood with a glucose meter (7). Blood glucose levels in Fig. 1C were measured at a fixed time window (1:00 to 2:00 PM) from free-feeding mice. For glucose tolerance tests, blood was sampled at 0, 15, 30, 60, and 120 min after glucose injection (intraperitoneal, 2 g/kg body weight) from mice that were fasted for 14 h. For insulin tolerance tests, blood was sampled at 0, 15, 30, 60, and 120 min after injection of Novolin R (intraperitoneal, 0.75 units/kg body weight, Novo Nordisk Inc.) from the mice fasted for ∼6 h. For metabolic stress tests, mice were fed a high-fat diet (D12331, 58 kcal% fat with sucrose, Research Diets Inc.) for 3 mo.

Islet β Cell Mass Assay.

Islet cell mass and islet size were measured using whole-pancreas large-scale imaging and subsequent semiautomated analysis as shown before (66, 67). Briefly, 5-μm pancreas sections were immunostained with antibodies (all 1:500): polyclonal guinea pig anti-porcine insulin (Dako), mouse monoclonal anti-human glucagon (Sigma-Aldrich), polyclonal goat anti-somatostatin (Santa Cruz), and DAPI, followed by the secondary antibody conjugated with DyLight 488, 549, and 649 (1:200, Jackson ImmunoResearch Laboratory). Images were taken using the Olympus IX8 DSU spinning disk confocal microscope and analyzed using custom-written scripts for Fiji/ImageJ and MATLAB (MathWorks) and Mathematica (Wolfram Research).

Islet Isolation and β Cell Cultures.

Pancreatic islets of Langerhans and individual β cells were isolated and cultured as we previously described (7, 118) with minor changes. Islets were collected manually under a stereomicroscope from the enzyme-digested pancreas (10 min at 37 °C) after injecting collagenase-P (0.25 mg/mL, 2 mL, Roche) through the common bile duct. Purified islets were dispersed into single cells by gentle pipetting after Dispase-II/EGTA (ethylene glycol-bis(2-aminoethylether)-N,N,N',N'-tetraacetic acid) (0.8 U/mL, 2 mM) incubation for 5 to 10 min. Individual cells were plated on coverslips precoated with 0.1% poly-D-lysine and grew in DMEM (Dulbecco's Modified Eagle Medium, Gibco) + 10% FBS (fetal bovine serum) culture medium with 5 mM glucose.

Cell Membrane Capacitance Measurements.

Capacitance recordings were performed under the whole-cell patch-clamp mode from β cell cultures as we described before (7). β cells were maintained at 34 to 36 °C using a heated chamber (RC-26GLP with PH-1 heating plate, Warner Instruments) fixed on an IX-70 inverted microscope (Olympus). EPC10-2 amplifier is under the control of Patch-Master (HEKA Electronik) (47, 118). The glass pipettes were 3 to 5 MΩ before recordings, holding membrane potential (V_h) was −70 mV, and the series resistance (Rs) was compensated >80% during recordings. β cells were identified by their characteristic sodium channels before recordings, as described in detail previously (7). These channels are inactivated at −70 mV but activated at −120 mV (79, 118). C_m was measured with “Sine + DC (direct current)” technique via the lock-in extension of Patch-Master (47), and sinewaves were 1 kHz and 60 mV peak–peak amplitude. Exocytosis was calculated as the C_m increase right after depolarization, and endocytosis as the C_m reduction, whose decay was fitted with a double exponential function. Endocytosis during a train of pulses was analyzed as described previously (47). Individual C_m traces from each genotype were pooled and averaged for analysis. All data were analyzed with Patch-Master and Igor Pro (Wavemetrics). Extracellular solution contained (in millimolars) 118 NaCl, 20 TEA-Cl, 5.6 KCl, 1.2 MgCl2, 2.6 CaCl2, 10 Hepes, and 5 d-glucose (pH = 7.4, ∼298 mOsm). The pipette solution contained (in millimolars) 118 Cs-glutamate, 20 TEA-Cl, 10 NaCl, 1 MgCl2, 0.05 CsEGTA, 3 Mg-ATP, 0.35 GTP, 0.1 cyclic-AMP, and 10 Hepes (pH 7.2 with CsOH, ∼295 mOsm).

Transferrin Uptake Assay.

Transferrin uptake assay is performed as described before (7). Cultured β cells were incubated with transferrin-AF567 (25 µg/mL, Invitrogen, T23365) for 10 min in culture medium (37 °C) after 2 h preincubation in serum-free DMEM, then washed twice with NaCl acidic buffer (pH = 5.3 with acetic acid, 2 min each) and fixed with 2% PFA. Cells were stained with anti-insulin and imaged under the confocal microscope at the middle level of β cells. The fluorescence intensity of each cell was measured in NIS-Elements from 6 to 10 cells per group in each experiment, and a total of 3 to 5 independent experiments were performed. For cell surface transferrin assay (Fig. 3F), the acid-wash step was skipped.

EM and Tomography.

Transmission EM was carried out as described previously (7, 47). In brief, purified pancreatic islets were fixed with 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer for 2 h and postfixed in 1% OsO₄, 1.5% K₄Fe(CN)₆, 0.1 M sodium cacodylate for 1 h. Islets were en bloc stained, dehydrated, embedded, and cut into ultrathin sections (70 to 90 nm), followed by imaging with an electron microscope. For EM tomography, tissue sections (200 to 250 nm) were imaged with a TECNAI TF20 transmission electron microscope (FEI) at 200 kV (Yale Center for Cellular and Molecular Imaging, Yale University) as described previously (75), and contours of membranes were traced using IMOD software manually. For EM quantification, at least two mice per genotype per condition were examined, and multiple EM sections from random regions in islets were collected. Granule docking (Fig. 6) was defined visually by the presence of direct granule-PM contact. Granule distribution was analyzed by the distance between the PM and each granule (at the closest point on the granule membrane).

Islet Perifusion and Insulin Secretion Assays.

Islet perifusion was performed at 37 °C as described previously (7). Briefly, purified islets of equal number (∼30) and size were paired between control and TKO in each experiment. Islets were perfused for 30 min with Krebs-Ringer bicarbonate buffer (KRBB) bubbled with 95% O₂ and 5% CO₂, and fractions were collected in 96-well plates at a 1-min interval. Islets were perfused with KRBB containing 2.8 mM for 10 min as a baseline and 20 mM glucose for 30 min, followed by 2.8 mM glucose for 10 min and 30 mM KCl perfusion in the end. Insulin was measured with insulin AlphaLISA (#AL204C, 2 μL scale, PerkinElmer Inc.). For the area under curve analysis, the first and second phases were estimated from 0 to 10 min and 10 to 30 min after the glucose stimulation. KRBB contained (in millimolars) 120 NaCl, 4.7 KCl, 2.5 CaCl₂, 1.2 MgSO₄, 25 NaHCO₃, 1.2 KH2PO₄, 10 Hepes, 2.8 d-glucose, and 0.1% BSA (bovine serum albumin).

Intracellular Ca²⁺ Concentration ([Ca²⁺]_i) and NADPH Imaging.

[Ca²⁺]_i of islet β cells were measured using the TILL-Photonics imaging system at 34 °C, as previously described (7). Purified islets with a medium size (100 to 200 μm in diameter) were preloaded with 5 μM Fura-2 acetoxymethyl ester (TEFLabs) for 30 min. Islet emission fluorescence under a sequential 350 and 380 nm light was collected by a water immersion lens (20×, 1.0 NA) and recorded with an EMCCD camera (iXon DU885, Andor, 4 × 4 binning, 5 ms) at 0.5 Hz. [Ca²⁺]_i from individual islets was quantified after in vitro calibration (7, 72). A total of 10 to 20 islets per experiment were used, and each experiment was repeated two to three times. KRBB containing 2.8 mM glucose, 12.8 mM glucose, or 30 mM KCl was perfused to the islets.

NADPH fluorescence imaging was acquired with the same imaging system (7). Islets were excited at 372 nm and imaged at 0.5 Hz through an ET525/50 filter (Chroma) since it elicited maximal NADPH fluorescence in our system. Fluorescence intensity from each islet was used to estimate NADPH changes between control and TKO islets. In addition, background fluorescence was subjected before analysis.

Statistics.

All values were presented as mean ± SEM unless otherwise specified. Statistic comparisons were conducted with a two-tailed Student’s t test for a two-sample comparison. ANOVA with the post hoc Tukey honest significant difference test was applied for three or more samples. P values of less than 0.05 were considered statistically significant.

Study Approval.

All procedures involving animals were approved by the Institutional Animal Care and Use Committee of the Medical College of Wisconsin and followed the NIH Guide for the Care and Use of Laboratory Animals (119).

Data Availability

All study data are included in the article and/or SI Appendix.