In Vivo Deletion of β-Cell Drp1 Impairs Insulin Secretion Without Affecting Islet Oxygen Consumption

Abstract

Mitochondria are dynamic organelles that undergo frequent fission and fusion events. Mitochondrial fission is required for ATP production, the tricarboxylic acid cycle, and processes beyond metabolism in a cell-type specific manner. Ex vivo and cell line studies have demonstrated that Drp1, a central regulator of mitochondrial fission, is required for glucose-stimulated insulin secretion (GSIS) in pancreatic β cells. Herein, we set out to interrogate the role of Drp1 in β-cell insulin secretion in vivo. We generated β-cell–specific Drp1 knockout (KO) mice (Drp1β-KO) by crossing a conditional allele of Drp1 to Ins1^cre mice, in which Cre recombinase replaces the coding region of the Ins1 gene. Drp1β-KO mice were glucose intolerant due to impaired GSIS but did not progress to fasting hyperglycemia as adults. Despite markedly abnormal mitochondrial morphology, Drp1β-KO islets exhibited normal oxygen consumption rates and an unchanged glucose threshold for intracellular calcium mobilization. Instead, the most profound consequences of β-cell Drp1 deletion were impaired second-phase insulin secretion and impaired glucose-stimulated amplification of insulin secretion. Our data establish Drp1 as an important regulator of insulin secretion in vivo and demonstrate a role for Drp1 in metabolic amplification and calcium handling without affecting oxygen consumption.

Islets from mice with β-cell–specific deletion of Drp1, a mitochondrial fission protein, have normal oxygen consumption but impaired second-phase and metabolic amplification of insulin secretion.

Glucose-stimulated insulin secretion (GSIS) from pancreatic β cells comprises interdependent pathways coupling glucose metabolism to insulin vesicle exocytosis (1–3). In the triggering pathway, glucose metabolism produces an increase in the cytosolic ATP/ADP ratio, which triggers the closure of ATP-sensitive K⁺ channels (K_ATP channels) and the opening of voltage-gated Ca²⁺ channels. The resulting influx of Ca²⁺ signals the exocytotic machinery to initiate insulin vesicle fusion with the plasma membrane. The metabolic amplifying pathways (previously termed the K_ATP-independent pathways) are responsible for the glucose-dependent increase in insulin secretion that occurs when cytosolic Ca²⁺ is already elevated (1–3). As mitochondria are central to both the triggering and amplifying pathways, healthy β-cell mitochondria are of upmost importance to glucose homeostasis (4).

The interdependent processes of mitochondrial fusion and fission represent an important, albeit incompletely understood, level of metabolic regulation. Disrupting the balance between fission and fusion has been shown to affect the tricarboxylic acid cycle, oxygen consumption, ATP and reactive oxygen species production, and mitochondrial processes beyond metabolism in various cell types (5). Mitochondrial fission and fusion are themselves regulated by diverse stimuli, including cellular energy balance, proliferation, and mitochondrial function, among others, thus establishing a dynamic bidirectional feedback loop between cellular metabolism and mitochondrial morphology (5).

Previous studies examining the in vivo consequences of impairing β-cell mitochondrial fusion, either directly through the knockout (KO) of the fusion protein Opa1 (6) or indirectly through deletion of prohibitin (7), reported profound mitochondrial defects that impaired insulin secretion through defective glucose-stimulated ATP production. Ex vivo, knockdown (8–10), or chemical inhibition of Drp1, a dynamin-related GTPase required for mitochondrial fission (10), have also been found to impair GSIS and ATP-linked oxygen consumption. However, these ex vivo studies may be confounded by the recently discovered off-target effect of the Drp1 inhibitor mDivi-1 on complex I of the electron transport chain (ETC) (11). Thus, it is not clear whether the consequences of blocking β-cell mitochondrial fission have been fully realized, especially because in vivo studies have not been performed.

Here, we generated a mouse model in which Drp1 is selectively deleted in pancreatic β cells (Drp1β-KO mice). We show that Drp1β-KO mice are glucose intolerant due to impaired GSIS and, despite markedly abnormal mitochondrial morphology, Drp1β-KO islets exhibit normal oxygen consumption rates (OCRs) and an unchanged glucose threshold for intracellular calcium mobilization. Instead, the most profound consequence of β-cell Drp1 deletion appears to be the impaired glucose-stimulated amplification of insulin secretion.

Materials and Methods

Animals

Dnm1l^tm1.1hise (MGI 4366510) and Ins1^{tm1.1(cre)Thor} (MGI 5614359) alleles have been described previously (12, 13) and are hereafter referred to as Drp1^f and Ins1^cre, respectively. Mice were maintained on a C57Bl/6 background. All mice were verified to be homozygous for the C57Bl/6J allele of nicotinamide nucleotide transferase (14). The following abbreviations were used for mouse genotypes: Drp1β-KO (Ins1^cre/+;Drp1^f/f), Drp1β-Het (Ins1^cre/+;Drp1^f/+), Drp1β–wild-type (WT) (Ins1^cre/+;Drp1^+/+, Ins1^+/+;Drp1^+/+, Ins1^+/+;Drp1^f/+, Ins1^+/+;Drp1^f/f). Drp1β-WT mice included Ins1^+/+ mice of any Drp1 genotype, because the presence of a single copy of the Ins1^cre allele did not affect glucose tolerance on a chow diet (Supplemental Fig. 1). Mice were group-housed in a colony maintained with a standard 12-hour light–dark cycle and given food and water ad libitum. Experiments were conducted according to the Guide for the Care and Use of Laboratory Animals, as adopted by the National Institutes of Health, and with approval of the University of California, San Francisco, Institutional Animal Care and Use Committee. For IP glucose tolerance tests, mice were individually housed and fasted for 6 hours before injection of glucose (2 mg/g body mass), after which tail-vein blood glucose measurements were made at regular intervals. Plasma insulin samples were collected before and 15 minutes after IP glucose injection and analyzed by ELISA (catalog no. 10-1247-10; Mercodia). For high-fat diet (HFD) experiments, 10-week-old female mice were fed Teklad Diet TD.07011 (4.9 Kcal/g, 21% from protein, 24.6% from carbohydrates, and 54.4% from fat; Envigo) for 12 weeks. For all other experiments, animals were fed PicoLab Mouse Diet 20 (3.75 Kcal/g, 23.2% from protein, 55.2% from carbohydrates, 21.6% from fat; Purina Mills). Pancreatic islet isolations were conducted according to a previously described protocol (15).

Immunofluorescence imaging

For imaging of mitochondrial morphology in individual β cells, islets were first rested overnight in islet media (RPMI with 10% fetal bovine serum, 1% penicillin/streptomycin, 25 mM HEPES) and then dissociated with 0.25% trypsin and plated onto glass coverslips coated with 804G cell-conditioned media. Two days after plating, cells were fixed in 4% paraformaldehyde for 10 minutes at 37°C, permeabilized in 0.1% Triton X-100 in PBS for 1 hour at room temperature (RT), blocked in 10% donkey serum for 1 hour at RT, and incubated in primary antibody overnight at 4°C in 1% donkey serum. Secondary antibodies were incubated for 1 hour at RT. Z-stacks were collected by confocal microscope at ×63 (Leica SP5; Leica Microsystems) for images of mitochondrial morphology using Tom20 immunostaining to accurately represent the three-dimensional nature of mitochondria. For peroxisomal morphology, cells were fixed as described but were permeabilized in 0.5% Triton X-100 in PBS for 15 minutes at RT, blocked in 2% BSA in PBS for 1 hour at RT, and incubated in primary antibody overnight in 2% BSA at 4°C. The following primary antibodies were used: rabbit anti-Tom20 [1:500; catalog no. sc11415; RRID: AB_2207533 (16); Santa Cruz Biotechnology]; rabbit anti-Pex14 [1:200; catalog no. 10594-1-AP; RRID: AB_2252194 (17); Proteintech Group]; guinea pig anti-insulin [1:500; catalog no. A0564, RRID: AB_10013624 (18); Dako].

For immunohistochemistry, pancreata were fixed in 4% paraformaldehyde and frozen in optimal cutting temperature compound. Tissue sections (6 µm) were permeabilized in 0.1% Triton X-100 and stained. The following primary antibodies were used: guinea pig anti-insulin [1:250; catalog no. A0564; RRID: AB_10013624 (19); Dako], rabbit anti-glucagon [1:250; catalog no. A0565; RRID: AB_10013726 (20); Dako]. The following secondary antibodies were used: anti–guinea pig Alexa Fluor 488 [1:500; catalog no. A11073; RRID: AB_142018 (21)], anti-rabbit Alexa Fluor 555 [1:500; catalog no. A31572; RRID: AB_162543 (22)], and goat anti-guinea pig Alexa 647 [1:500; catalog no. A21450; RRID: AB_141882 (23)], all from Life Technologies.

Western blots

Protein was extracted from pancreatic islets using radioimmune precipitation buffer containing complete protease inhibitors (Roche). Lysates were sonicated on ice for 5 minutes with alternating 30-second pulses and rests and precleared by centrifugation at 15,000g for 5 minutes at 4°C. Mini-Protean Tetra and Trans-Blot Turbo systems (Bio-Rad Laboratories) were used for PAGE gel electrophoresis and polyvinylidene fluoride membrane transfer, respectively. Band intensities were quantified using ImageJ software (https://imagej.nih.gov/ij/) (24). Primary antibodies used were mouse anti-DLP1 [1:1000; catalog no. 61113; RRID: AB_398424 (25); BD Biosciences], mouse anti-GAPDH-peroxidase [1:20,000; catalog no. G9295; RRID: AB_1078992 (26); Sigma-Aldrich], mouse anti-rodent oxidative phosphorylation [1:1000; catalog no. ab110413; RRID: AB_2629281 (27); Abcam], rabbit anti-vinculin [1:1000; catalog no. 13901; RRID: AB_2728768 (28); Cell Signaling Technology], rabbit anti-Opa1 [1:1000; catalog no. 80471; RRID: AB_2734117 (29); Cell Signaling Technology], and the following from Proteintech Group: rabbit anti-Fis1 [1:1000; catalog no. 10956-1-AP; RRID: AB_2102532 (30)], rabbit anti-Mff [1:1000; catalog no. 17090-1-AP; RRID: AB_2142463 (31)], rabbit anti-Mid51 [1:500; catalog no. 20164-1-AP; RRID: AB_10639522 (32)], rabbit anti-Mfn1 [1:1000; catalog no. 13798-1-AP; RRID: AB_2266318 (33)], and rabbit anti-Mfn2 [1:1000; catalog no. 12186-1-AP; RRID: AB_2266320 (34)]. Secondary antibodies used were horse anti-mouse horseradish peroxidase [1:10,000; catalog no. 7076; RRID: AB_330924 (35)] and goat anti-rabbit horseradish peroxidase [1:10,000; catalog no. 7074; RRID: AB_2099233 (36)], both from Cell Signaling Technology.

GSIS

Islets were rested for 24 hours after isolation in islet media (RPMI with 10% fetal bovine serum, 1% penicillin/streptomycin, and 25 mM HEPES). For static GSIS, 20 islets were hand picked and equilibrated in Krebs-Ringer bicarbonate HEPES buffer (KRBH; 137 mM NaCl, 4.7 mM KCl, 1.2 mM KH₂PO₄, 1.2 mM MgSO₄, 2.5 mM CaCl₂, 25 mM NaHCO₃, 20 mM HEPES, and 0.25% BSA) containing 2.8 mM glucose for 1 hour at 37°C. Equilibration buffer was then removed and islets were sequentially stimulated with KRBH containing concentrations of glucose and compounds as indicated. Islets were lysed in radioimmune precipitation buffer for determination of total islet insulin content for stimulation-buffer insulin measurement normalization. Dynamic GSIS was carried out on a perifusion system (Biorep Technologies) according to the manufacturer’s protocol.

OCRs

Islet oxygen consumption measurements were conducted on a Seahorse XFe24 machine (Agilent Technologies). Islets were placed in the wells of an islet plate in XF Base Media (Agilent Technologies) with 3 mM glucose and equilibrated at 37°C for 1 hour in a CO₂-free incubator. Drug concentrations were as follows: oligomycin A (5 µM), rotenone (5 µM) antimycin A (5 µM). OCRs were normalized to islet DNA content.

Two-photon imaging of NADH and NADPH

NADH and NADPH [collectively termed NAD(P)H] images were collected as described by Gregg et al. (37). NAD(P)H was excited with Mai Tai DeepSee Ti:Sapphire laser (Spectra-Physics) at 740 nm with a 450/70m bandpass emission filter (Chroma Technology) before being collected by a H7422P-40 GaAsP photomultiplier tube (Hamamatsu Photonics). Images were collected with a Nikon Plan Apo ×60/1.4NA objective at 256 × 256 resolution with 120-second (1/s) collection using SPC-830 Photon Counting Electronics (Becker & Hickl). A single region of interest was used to quantify average islet NAD(P)H fluorescence intensity using Fiji (https://imagej.net/Fiji) (24).

Islet NADP⁺ and NADPH

Islets were rested for 24 hours after isolation in standard islet media (RPMI 1640, 10% FBS, 1% penicillin/streptomycin, and 125 mM HEPES). Islets were dispersed, plated, and rested for an additional 24 hours. NADP⁺ and NADPH were measured using the NADP/NADPH-Glo kit (catalog no. G9081; Promega) following a 15-minute challenge with either 2.8 mM KRBH or 16.7 mM KRBH.

Islet quantitative PCR

Islets were rested for 24 hours after isolation in standard islet media (RPMI 1640, 10% FBS, 1% penicillin/streptomycin, and 125 mM HEPES). RNA samples and cDNA libraries were prepared using DirectZol (Zymo Research) and Superscript III Reverse transcription (Thermo Fisher Scientific), respectively. Quantitative PCR analysis was carried out using a QuantStudio 5 Real-Time PCR system (Applied Biosystems). Slc2a2 and Gck expression in each islet was calculated using the comparative threshold cycle method relative to the average of Ppia and Actb. Primers used are as follows: Slc2a2 (mm00446229_m1); Gck (forward: GAGATGGATGTGGTGGCAAT, reverse: TCACATTGGCGGTCTTCATA, probe: GACACGGTGGCCACAAT); Ppia (forward: GGCCGATGACGAGCCC, reverse: TGTCTTTGGAACTTTGTCTGCAA, probe: TGGGCCGCGTCTCCTTCGA); Actb (forward: TTCAACACCCCAGCCATGTA, reverse: TGTGGTACGACCAGAGGCATAC, probe: TAGCCATCCAGGCTGTGCTGTCCC).

Time-lapse imaging of cytosolic islet Ca²⁺

Imaging conditions matched those described in Gregg et al. (37), with an added islet labeling protocol that allowed for simultaneous imaging of Drp1β-WT and Drp1β-KO islets. Islets were labeled for 10 minutes at 37°C in media containing 1.25 µg/mL DiOC₁₈(7) (diR; catalog no. D-12731; Molecular Probes) prepared from a concentrated stock of 1 mg/mL in dimethyl sulfoxide (stored at 4°C), followed by preincubation in 2.5 µM FuraRed AM (catalog no. F3020; Molecular Probes) for 45 minutes at 37°C. The islets were then transferred to the imaging chamber where a single diR image was taken using a Cy7 cube (Exciter, ET710/75x; dichroic, T760lpxr, Emitter, ET810/90m; Chroma Technology). We confirmed that the presence diR had no effect on Ca²⁺ oscillations in WT mice. Ca²⁺ was imaged using a triple-edge standard epi-fluorescence dichroic beamsplitter (catalog no. FF444/521/608-Di01; Semrock) in combination with Chroma ET430/20x and ET500/20x exciters and an ET630/70m emitter. The FuraRed ratio was reported as the 430/500 excitation ratio normalized to signal in the presence of 5 mM KCN. The solution flow rate was maintained at 0.35 mL/min using an MCFS-EZ microfluidic flow control system (Fluigent), and temperature was maintained at 33°C using inline and solution heaters (Warner Instruments). Fluorescence emission was collected with an ORCA-Flash4.0 V2 Digital CMOS camera (Hamamatsu Photonics) with a sampling interval of 6 seconds. A single region of interest was used to quantify the average response of each islet, using Nikon Elements (Nikon Instruments) and MATLAB software (MathWorks).

Statistical analysis

Data are presented as mean ± SEM. Statistical tests were completed using Prism (GraphPad Software) as described in the figure legends. The Holm-Sidak multiple comparisons test was used to make comparisons between group means in two-way ANOVA analyses with three groups. For one- and two-way ANOVA analyses, the results of Bonferroni corrected unpaired t tests (or other test as indicated in the figure legends) were used to make Drp1β-WT vs Drp1β-KO and Drp1β-WT vs Drp1β-Het comparisons at individual time points, and these are displayed in figure panels.

Results

Drp1β-KO mice were glucose intolerant

To generate mice selectively lacking β cell Drp1, we crossed mice carrying a conditional allele of Drp1 (13) with Ins1^cre mice in which Cre recombinase replaces the coding region of the Ins1 gene (12). We thus generated Drp1β-WT, Drp1β-Het, and Drp1β-KO mice, as defined in Methods. Drp1 protein was nearly undetectable in Drp1β-KO islets and was reduced by ∼50% in Drp1β-Het islets (Fig. 1A and 1B). As in other models of Drp1 knockdown (8–10), we found that Drp1β-KO β cells contained a large perinuclear mitochondrial structure, with occasional linear and circular protrusions, consistent with a superfused mitochondrial network (Fig. 1C; Supplemental Fig. 2). In contrast, Drp1β-WT mitochondria were networked diffusely throughout the cytoplasm, whereas Drp1β-Het β-cell mitochondria largely resembled those of WT controls. Levels of other mitochondrial fission and fusion proteins were unchanged (Supplemental Fig. 3A–3G). Drp1β-KO β cells also displayed elongated peroxisomes compared with the spherical peroxisomes of Drp1β-WT β cells (Supplemental Fig. 3H), consistent with the known role of Drp1 in peroxisomal fission (38–40).

Figure 1.

Drp1β-KO mice are glucose intolerant. (A) Western blot for Drp1 and vinculin (loading control) in male, 17- to 18-week-old Drp1β-WT, Drp1β-Het, and Drp1β-KO islets. (B) Drp1 signal from the Western blot results in islets in (A) quantified and normalized to vinculin and WT islets [one-way ANOVA, F_{(2, 6)} = 19.91; P = 0.0022]. (C) Immunofluorescence images of dispersed 13- to 22-week-old islets stained for insulin (green), Tom20 (white), and DAPI (blue). Scale bar is 10 µm. (D) IPGTT results in male, 8- to 12-week-old Drp1β-WT (blue circles), -Het (green squares), and -KO (red triangles) mice (n = 8 to 12 per genotype). Two-way ANOVA with repeated measures data are as follows: interaction F_{(10, 135)} = 21.95, P < 0.0001; time F_{(5, 135)} = 266.2, P < 0.0001; genotype F_{(2, 27)} = 72.34, P < 0.0001 (Drp1β-WT vs Drp1β-Het, P = 0.1939; Drp1β-WT vs Drp1β-KO, P < 0.0001). (E) AUC analysis of the IPGTT results in (D). One-way ANOVA data are as follows: F_{(2, 27)} = 87.65, P < 0.0001. (F) IPGTT results in female 8- to 12-week-old Drp1β-WT, -Het, and -KO mice (n = 8 to 11 per genotype). Two-way ANOVA with repeated measures data are as follows: interaction F_{(10, 125)} = 10.92, P < 0.0001; time F_{(5, 125)} = 331.3, P < 0.0001; genotype F_{(2, 25)} = 16, P < 0.0001 (Drp1β-WT vs Drp1β-Het, P = 0.5881; Drp1β-WT vs Drp1β-KO, P < 0.0001). (G) AUC analysis of the IPGTT results in (F). One-way ANOVA data are as follows: F_{(2, 25)} = 18.89, P < 0.0001. (H) IPGTT results in female, 30-week-old, Drp1β-WT, -Het, and -KO mice (n = 5 per genotype). Red dotted line represents maximum measurement of 500 mg/dL. Two-way ANOVA with repeated measures data are as follows: interaction F_{(10, 60)} = 23.44, P < 0.0001; time F_{(5, 60)} = 149, P < 0.0001; genotype F_{(2, 12)} = 23.72, P < 0.0001 (Drp1β-WT vs Drp1β-Het, P = 0.3684; Drp1β-WT vs Drp1β-KO, P < 0.0001). (I) AUC analysis of IPGTT results in (H). One-way ANOVA data are as follows: F_{(2, 12)} = 26.94, P < 0.0001. (J) Fasting blood glucose values in 16-, 24-, and 30-week-old female mice (n = 4 to 5 per genotype). Two-way ANOVA data are as follows: interaction F_{(4, 35)} = 0.02826, P = 0.9984; age F_{(2, 35)} = 3.504, P = 0.0410; genotype F_{(2, 35)} = 5.383; P = 0.0092 (Drp1β-WT vs Drp1β-Het, P = 0.0051; Drp1β-WT vs Drp1β-KO, P = 0.2413). *P ≤ 0.05; **, P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. AUC, area under the curve; BG, blood glucose; DAPI, 4′,6-diamidino-2-phenylindole; IPGTT, IP glucose tolerance test; ns, not significant.

At 8 to 12 weeks of age, male and female Drp1β-KO mice were hyperglycemic at 30, 60, 90, and 120 minutes after IP glucose injection and by area under the curve (AUC) analysis (Fig. 1D–1G), compared with Drp1β-WT mice. Drp1β-Het mice were not significantly different from Drp1β-WT mice by either method. At 30 weeks of age, female Drp1β-KO mice were more glucose intolerant than at 8 to 12 weeks (Fig. 1H and 1I). However, fasting blood glucose (Fig. 1J) and random blood glucose levels (Supplemental Fig. 4A), at ages up to 30 weeks, were not significantly changed. Insulin tolerance was also unchanged (Supplemental Fig. 4B). After 12 weeks of HFD (starting at 10 weeks of age), female Drp1β-KO mice were again glucose intolerant (Supplemental Fig. 4C and 4D) but still did not develop fasting hyperglycemia. HFD did not change glucose tolerance in Drp1β-Het mice compared with Drp1β-WT controls. We note, however, that the effect of the Ins1^cre allele on glucose tolerance in HFD conditions has not been determined. These results indicate that the loss of β-cell Drp1 results in glucose intolerance without fasting hyperglycemia.

Drp1β-KO islets exhibited impaired GSIS and mitochondrial abnormalities

To explore the cause of Drp1β-KO glucose intolerance, we assayed insulin secretion in vivo. Male Drp1β-KO mice had reduced plasma insulin concentrations 15 minutes after IP glucose injection compared with Drp1β-WT controls (Fig. 2A). Immunostaining for insulin and glucagon revealed no striking differences in islet architecture between Drp1β-KO and Drp1β-WT mice (Supplemental Fig. 4E). Ex vivo, male Drp1β-KO islets secreted less insulin in response to stimulation with 16.7 mM glucose than Drp1β-WT islets, with no change in response to 2.8 mM glucose (Fig. 2B). Secretion in response to KCl at 2.8 mM glucose was unchanged (Fig. 2C), suggesting that Drp1 deletion impaired GSIS upstream of plasma membrane depolarization. The insulin content of isolated Drp1β-KO islets was unchanged (Fig. 2D), indicating that the GSIS defect was not due to lack of insulin production. We thus hypothesized that the Drp1β-KO GSIS defect arose in the mitochondria.

Figure 2.

Drp1β-KO mice have impaired GSIS and mitochondrial abnormalities. (A) Plasma insulin measurements in 10-week-old, male Drp1β-WT and -KO mice (n = 17 to 18 per genotype). Two-way ANOVA with repeated measures data are as follows: interaction F_{(1, 33)} = 25.14, P < 0.0001; time F_{(1, 33)} = 11.31, P = 0.0020; genotype F(_{1, 33)} = 2.697, P = 0.1100. (B) Ex vivo GSIS of male, 8- to 17-week-old, Drp1β-WT and -KO islets in response to 2.8 mM and 16.7 mM glucose (n = 10 per genotype). Two-way ANOVA with repeated measures data are as follows: interaction F_{(1, 18)} = 8.214, P = 0.0103; treatment F_{(1, 18)} = 27.89, P < 0.0001; genotype F_{(1, 18)} = 6.429, P = 0.0207. (C) Ex vivo GSIS of male, 8- to 12-week-old, Drp1β-WT and -KO islets in response to 2.8 mM glucose and 30 mM KCl (n = 4 to 5 per genotype). Two-way ANOVA with repeated measures data are as follows: interaction F_{(1, 7)} = 0.5459, P = 0.4840; treatment F_{(1, 7)} = 83.24, P < 0.0001; genotype F_{(1, 7)} = 0.5852, P = 0.4693. (D) Insulin content from 7- to 12-week-old male islets (n = 9 to 10 per genotype); unpaired t test P = 0.5877. (E) Western blots of selected ETC complex proteins and GAPDH in 17- to 18-week-old male islets, quantified and normalized to GAPDH and average of WT bands in (F–J). (F) NDUFB8: One-way ANOVA data are F_{(2, 6)} = 11.26, P = 0.0093. (G) SDHB: One-way ANOVA data are F_{(2, 6)} = 15.93, P = 0.0040. (H) UQCRC2: One-way ANOVA data are F_{(2, 6)} = 4.252, P = 0.0708. (I) MTCO1: One-way ANOVA data are F_{(2, 6)} = 6.322, P = 0.0333. (J) ATP5A: One-way ANOVA data are F_{(2, 6)} = 2.045, P = 0.2103. (K) Two-photon NAD(P)H images from Drp1β-WT and -KO islets. (L) Quantified NAD(P)H intensity in 6- to 7-week-old male islets (n = 48 to 54 islets from five mice of each genotype). Two-way ANOVA data are as follows: interaction F_{(1, 143)} = 57.27, P < 0.0001; treatment F_{(1, 143)} = 842.1, P < 0.0001; genotype F_{(1, 143)} = 11.71, P = 0.0008. (M) Islet NADPH levels in dispersed 12- to 22-week-old, male, Drp1β-WT and -KO islets 15 minutes after exposure to 2.8 mM or 16.7 mM glucose (n = 9 replicates from three mice of each genotype). Two-way ANOVA data are as follows: interaction F_{(1, 16)} = 0.1408, P = 0.7125; treatment F_{(1, 16)} = 111.6, P < 0.0001; genotype F_{(1, 16)} = 0.081, P = 0.7796. For all panels, post hoc significance levels are represented as follows: *P ≤ 0.05; **P ≤0.01; ****P ≤ 0.0001. KO, knockout; ns, not significant.

Studies in cardiomyocytes have shown a correlation between Drp1 expression and ETC complex levels (41, 42). Indeed, we found that NDUFB8, SDHB, and ATP5A, components of ETC complexes I, II, and IV respectively, were reduced in Drp1β-KO islets (Fig. 2E–2J). UQCRC2 (complex III) was nonsignificantly reduced (P = 0.13), whereas MTCO1 (complex V) was unchanged by Drp1 deletion. We next used two-photon imaging of NAD(P)H to directly image β-cell metabolism (37, 43). As expected from the mitochondrial morphology images in Fig. 1C, the NAD(P)H fluorescence in live Drp1β-WT β cells was distributed throughout the cytoplasm, whereas the NAD(P)H fluorescence in Drp1β-KO β cells was clustered in a perinuclear location (Fig. 2K). Quantification of these NAD(P)H images revealed that Drp1β-KO islets exhibited elevated NAD(P)H fluorescence at 2 mM glucose and reduced NAD(P)H at 10 mM glucose when compared with Drp1β-WT islets (Fig. 2L). NADPH levels measured using a bioluminescence-based method were unchanged between Drp1β-WT and -KO islets 15 minutes after glucose stimulation (Fig. 2M). These observations raise the question of whether abnormal mitochondrial metabolism contributes to defective Drp1β-KO GSIS.

β-cell Drp1 deletion did not affect islet oxygen consumption or the glucose threshold for islet calcium entry

To ascertain if Drp1β-KO islet mitochondrial abnormalities translate into impaired flux through the ETC, we measured OCRs in intact Drp1β-KO islets. Surprisingly, oxygen consumption by Drp1β-KO islets was indistinguishable from that of Drp1β-WT islets in response to sequential stimulation with 9 mM glucose, 16.7 mM glucose, and the mitochondrial poisons oligomycin and rotenone/antimycin A (Fig. 3A). As such, there were no significant changes in ATP-linked respiration (Fig. 3B), basal respiration (Fig. 3C), proton leak (Fig. 3D), or coupling efficiency (Fig. 3E). In contrast to the upregulation of Slc2a2 mRNA observed in Drp1-knockdown MIN6 cells (10), we found that mRNA expression of Glut2 (Slc2a2) and Gck was unchanged in Drp1β-KO islets (Fig. 3F). These data indicate that, surprisingly, respiration in Drp1β-KO islets is normal despite lower ETC component expression and abnormal NAD(P)H levels.

Figure 3.

β-cell Drp1 deletion does not affect islet oxygen consumption or the glucose threshold for Ca²⁺ entry. (A) OCRs in islets from 8- to 11-week-old male Drp1β-WT (blue circles) and -KO (red triangles) mice (n = 12 to 18 from four mice of each genotype). Two-way ANOVA with repeated measures data are as follows: interaction F_{(37, 1036)} = 0.8141, P = 0.7785; time F_{(37, 1036)} = 163.9, P < 0.0001; genotype F_{(1, 28)} = 0.0002147, P = 0.9884. (B–E) Respiration parameters calculated from OCRs in (A). (B) ATP-linked respiration unpaired t test, P = 0.2223. (C) Basal respiration unpaired t test, P = 0.8950. (D) Proton leak unpaired t test, P = 0.7929. (E) Coupling efficiency unpaired t test, P = 0.2015. (F) Quantitative PCR measurement of islet Glut2 (Slc2a2) and Gck expression in 11- to 16-week-old Drp1β-WT and -KO islets (n = 3 to 4 mice of each genotype). Unpaired t tests: Slc2a2, P > 0.99; Gck, P = 0.1042. (G) Average cytosolic Ca²⁺ in male 11- to 17-week-old islets in response to escalating glucose concentrations, normalized to 5 mM KCN treatment (n = 115 to 131 islets from six mice of each genotype). Two-way ANOVA data are as follows: interaction F_{(13, 3416)} = 8.437, P < 0.0001; time F_{(13, 3416)} = 43.9, P < 0.0001; genotype F_{(1, 3416)} = 0.7382, P = 0.3903. *Results of Holm-Sidak corrected unpaired t tests. (H) Example islet Ca²⁺ oscillations in 5, 7, and 9 mM glucose. (I) Period calculations from islet Ca²⁺ traces in 9 mM glucose (n = 152 to 154 islets from six mice of each genotype). Unpaired t test P < 0.0001. (J) Amplitude calculation from islet Ca²⁺ traces in 9 mM glucose (n = 152 to 154 islets from six mice of each genotype). Unpaired t test P = 0.0005. (K) Duty cycle calculation from islet Ca²⁺ traces in 9 mM glucose. Unpaired t test P = 0.2533 (n = 152 to 154 islets from six mice of each genotype). *P ≤ 0.05; **P ≤0.01; ***P ≤ 0.001; ****P ≤ 0.0001. KO, knockout; ns, not significant; qPCR, quantitative PCR; Rot, rotenone.

Downstream of metabolism-dependent K_ATP channel closure, an acute spike in cytosolic Ca²⁺ is required to trigger insulin secretion, and persistently elevated cytosolic Ca²⁺ is necessary to sustain secretion. Normal oxygen consumption in Drp1β-KO islets would be expected to produce a pattern of cytosolic Ca²⁺ elevation that is similar to that of Drp1β-WT islets. Therefore, we imaged islet cytosolic Ca²⁺ influx across a range of glucose concentrations (2 to 9 mM; Fig. 3G) in Drp1β-WT and Drp1β-KO islets. Average cytosolic Ca²⁺ was elevated at 2 mM and 5 mM glucose in Drp1β-KO islets, but Drp1β-WT and -KO islets showed a robust “phase 0” in response to 5 mM glucose, reflecting Ca²⁺ uptake by the endoplasmic reticulum (44). Drp1β-WT and -KO islets exhibited an acute increase in cytosolic Ca²⁺ in response to 7 mM glucose, indicating the glucose threshold required to trigger cytosolic Ca²⁺ influx was unchanged by Drp1 deletion. Likewise, islets of both genotypes exhibited a spike in cytosolic Ca²⁺ upon exposure to 9 mM glucose. However, at 9 mM glucose, average cytosolic Ca²⁺ in Drp1β-KO islets then decreased to a level lower than that of Drp1β-WT.

On the level of the individual islet, oscillations in cytosolic Ca²⁺ during glucose exposure cause pulses of insulin secretion (45). Individual Drp1β-WT and -KO islet Ca²⁺ traces showed aperiodic depolarizations at 7 mM glucose and exhibited robust oscillations at 9 mM glucose (Fig. 3H). Surprisingly, the Ca²⁺ oscillation period was nearly doubled in Drp1β-KO islets (Fig. 3I), with a concomitant reduction in oscillation amplitude, calculated as the difference between Ca²⁺ peak and nadir (Fig. 3J). However, the duty cycle, calculated as the fraction of time that an individual islet spends with cytosolic Ca²⁺ levels above half maximal, was unchanged in Drp1β-KO islets at 9 mM glucose (Fig. 3K), again consistent with a normal threshold for glucose-stimulated cytosolic calcium increase (37, 45). These results show that Drp1β-KO islets have abnormal Ca²⁺ handling.

Loss of β cell Drp1 impaired second-phase insulin secretion and metabolic amplification

To look more closely at the effects of Drp1 deletion on insulin secretion, we measured insulin secretion dynamically, using an islet perifusion assay. Although first-phase insulin secretion was indistinguishable between Drp1β-WT and Drp1β-KO islets, second-phase insulin secretion was clearly impaired in Drp1β-KO islets (Fig. 4A). KCl-induced secretion at 2.8 mM glucose was also unchanged, consistent with our results from static GSIS assays (Fig. 2C). AUC analysis of the perifusion curve confirmed these findings: Total AUC and AUC during second phase were reduced in Drp1β-KO islets (Fig. 4B and 4D), whereas first phase AUC (Fig. 4C) and KCl-induced AUC (Fig. 4E) were unchanged.

Figure 4.

Loss of β-cell Drp1 impairs second-phase insulin secretion and the metabolic amplifying pathway. (A) Dynamic insulin secretion in female 7- to 12-week-old Drp1β-WT and -KO islets (n = 16 from eight mice of each genotype). Two-way ANOVA with repeated measures data are as follows: interaction F_{(24, 720)} = 2.151, P = 0.0012; time F_{(24, 720)} = 107.8, P < 0.0001; genotype F_{(1, 30)} = 6.709, P = 0.0147. (B–E) AUC analyses of (A). (B) Total AUC unpaired t test P = 0.0035. (C) First-phase AUC, representing secretion from minutes 20 to 24. Unpaired t test P = 0.2380. (D) Second-phase AUC, representing minutes 25 to 50. Unpaired t test P = 0.0009. (E) KCl AUC, representing secretion minutes 60 to 63. Unpaired t test P = 0.1461. (F) In vitro GSIS of male 7- to 9-week-old Drp1β-WT and -KO islets in response to 2.8 mM or 16.7 mM glucose in the presence of 200 mM Dz and 30 mM KCl (n = 6 of each genotype). Two-way ANOVA with repeated measures data are as follows: interaction F_{(1, 10)} = 1.736, P = 0.2170; treatment F_{(1, 10)} = 47.81, P < 0.0001; genotype F_{(1, 10)} = 8.537, P = 0.0153. *P ≤ 0.05; **P ≤0.01; ***P ≤ 0.001. Dz, diazoxide; KO, knockout.

Independent of ATP production and GSIS triggering, a number of glucose-derived mitochondrial metabolites influence insulin secretion through the metabolic amplifying pathway (45). We measured the metabolic amplification of insulin secretion in Drp1β-WT and Drp1β-KO islets by treating them with either 2.8 mM or 16.7 mM glucose in the presence of KCl (30 mM) and the K_ATP channel activator diazoxide (200 μM) (1–3). At 2.8 mM glucose, there was no difference between the genotypes, but at 16.7 mM glucose, Drp1β-KO islets displayed reduced insulin secretion compared with Drp1β-WT islets (Fig. 4F), reflecting impaired metabolic amplification.

Discussion

Herein, we have described an in vivo knockout of a mitochondrial fission protein in the pancreatic β cell. Our study corroborates previous in vitro studies that identified a reduction in GSIS upon loss of β-cell Drp1 and expands the known roles of β-cell Drp1 with our findings of impaired second-phase insulin secretion, abnormal islet Ca²⁺ handling, and defective metabolic amplification in Drp1β-KO islets.

We found that Drp1β-KO islet mitochondria were abnormal in several ways, exhibiting a highly fused morphology, decreased expression of ETC complexes, and altered NAD(P)H fluorescence at both low and high glucose concentrations. Nonetheless, Drp1β-KO islet oxygen consumption was unchanged, suggesting that Drp1β-KO islets retain adequate ETC expression to maintain normal mitochondrial oxidative phosphorylation and highlighting the pitfall of using expression level to infer enzymatic activity (46). Normal oxygen consumption is consistent with our unchanged threshold for glucose-stimulated cytosolic Ca²⁺ increase (Fig. 3G) and duty cycle measurements in Drp1β-KO islets (Fig. 3K). There is precedence for these observations: Drp1 deletion in primary fibroblasts and macrophages also does not impair oxygen consumption (47–49). Importantly, our data do show that there is a triggering pathway defect in Drp1β-KO islets, just not at the level of the ETC. Average Ca²⁺ levels are elevated at 2 mM and 5 mM glucose and reduced at 9 mM glucose after a normal initial rise. These changes in cytosolic Ca²⁺ could represent accommodation to a narrowed functional range of the triggering pathway components or a primary defect in Ca²⁺ handling caused by Drp1 loss. Although there are no reported links between elongated peroxisomes and insulin secretion, given the parallel role of Drp1 in peroxisomal fission (38–40), we also cannot rule out nonmitochondrial contributions to the Drp1β-KO GSIS defect. Studies are needed to distinguish among these possibilities.

In addition to clarifying the role of Drp1 in the triggering of insulin secretion, we have identified a link between Drp1 and the amplifying pathway of insulin secretion. Although the exact molecular mediator of the amplifying pathway that is deficient remains to be identified, a number of amplifying factors could feasibly be altered by loss of β-cell Drp1. Two attractive possibilities are either disruption of the isocitrate-to-SENP1 pathway, whereby glucose induces an elevation in cytosolic NADPH levels that activate the protease SENP1 and, in turn, the exocytotic machinery (50), or impaired mitochondrial GTP production (51). Indeed, two-photon imaging revealed decreased NAD(P)H fluorescence in Drp1β-KO islets at 10 mM glucose (Fig. 2L). However, our unchanged bioluminescence-based measurement of total cellular NADPH suggests that any alterations of the isocitrate-to-SENP1 pathway caused by Drp1 loss are not due to a complete block in glucose-induced NADPH production. Further defining the effect of Drp1 loss on β-cell metabolism will be an important future direction of our work.

There are several important differences between our data and those of previous studies examining β-cell Drp1 knockdown in vitro. Past studies using Drp1 knockdown or dominant negative Drp1 overexpression in β-cell lines or mDivi-1–mediated Drp1 inhibition in primary mouse islets have reported decreased ATP-linked respiration (9, 10) and/or increased proton leak (9, 10), in contrast to our findings of normal oxygen consumption in Drp1β-KO islets. This discrepancy could reflect a difference between proliferative β-cell lines and quiescent adult mouse islets (52), a difference in the consequences of acute in vitro vs chronic in vivo Drp1 loss, or the recently identified off-target effect of mDivi-1 on ETC complex I (11). Furthermore, in another study (10), researchers have suggested that impaired GSIS in mDivi-1–treated islets was caused by decreased substrate availability because it could be rescued by methyl pyruvate treatment. However, we note that methyl pyruvate may directly close K_ATP channels instead of acting as a mitochondrial substrate (53), making this rescue difficult to interpret. Our observations of unchanged Drp1β-KO islet respiration (Fig. 3A) indicate that even if such an effect on substrate availability exists, it is not sufficient to alter islet respiration and subsequent ATP production, at least in the setting of chronic Drp1 loss. Additional studies are needed to investigate the in vitro vs in vivo and acute vs chronic differences observed herein.

Finally, it is noteworthy that Drp1β-Het mice did not exhibit a glucose tolerance phenotype under either normal (Fig. 1) or HFD conditions (Supplemental Fig. 4), despite ∼50% reduced Drp1 levels. With the exception of several studies in cardiomyocytes (54, 55), Drp1 heterozygosity has not been found to produce a phenotype in most cell types (56). Our findings further illustrate the highly cell-type specific requirements for mitochondrial dynamics (57).

In summary, our findings establish an important role for β-cell Drp1 in regulating the second phase of GSIS, islet Ca²⁺ handling, and metabolic amplification, and suggest that a greater understanding of the role of mitochondrial fission in β cells may lead to novel approaches to correct islet dysfunction in diabetes.

Glossary

Abbreviations:

AUC

area under the curve

diR

DiOC₁₈(7)

ETC

electron transport chain

GSIS

glucose-stimulated insulin secretion

HFD

high-fat diet

K_ATP channel

ATP-sensitive K⁺ channel

KO

knockout

KRBH

Krebs-Ringer bicarbonate HEPES buffer

OCR

oxygen consumption rate

RT

room temperature

WT

wild-type