Mitofusins Mfn1 and Mfn2 Are Required to Preserve Glucose- but Not Incretin-Stimulated β-Cell Connectivity and Insulin Secretion

Mitochondria are often referred to as the powerhouses or “chief executive organelles” of the cell, using fuels to provide most of the energy required to sustain normal function (1). Mitochondrial oxidative metabolism plays a pivotal role in the response of pancreatic β-cells to stimulation by glucose and other nutrients (2). Thus, as blood glucose increases, enhanced glycolytic flux and oxidative metabolism lead to an increase in ATP synthesis, initiating a cascade of events that involve the closure of K_ATP channels (3), plasma membrane depolarization, and the influx of Ca²⁺ via voltage-dependent Ca²⁺ channels (VDCC). The latter, along with other less well-defined “amplifying” signals (4), drive the biphasic release of insulin (2). Gut-derived incretin hormones, including glucagon-like peptide-1 (GLP-1) and glucose-dependent insulinotropic peptide (GIP) (5), further potentiate secretion by binding to class-B G protein-coupled receptors (GPCRs) to generate cAMP and other intracellular signals (5).

Under normal physiological conditions, mitochondria undergo fusion and fission cycles that are essential for quality control and adaptation to energetic demands (6). Thus, highly interconnected mitochondrial networks allow communication and interchange of contents between mitochondrial compartments as well as with other organelles such as the endoplasmic reticulum (ER) (7). These networks exist interchangeably with more fragmented structures, displaying more “classical” mitochondrial morphology (8). Mitochondrial fission is also necessary for “quality control” and the elimination of damaged mitochondria by mitophagy (9).

While the mitofusins MFN1 and MFN2, homologs of the Drosophila melanogaster fuzzy onions (fzo) and mitofusin (dmfn) gene products (10), are guanosine-5'-triphosphatases that mediate fusion of the outer mitochondrial membrane, optic atrophy protein 1 (OPA1) controls that of the inner mitochondrial membrane. Dynamin-related protein 1 (DRP1) is responsible for mitochondrial fission (11). Other regulators include mitochondrial fission 1 protein (FIS1), mitochondrial fission factor (MFF), and MiD49/51 (12).

Earlier studies (13–18) have shown that perturbations in mitochondrial structure in β-cells have marked effects on glucose-stimulated insulin secretion (GSIS). Surprisingly, whether the canonical and evolutionarily conserved machinery involved in mitochondrial fusion (i.e., the mitofusins), control mitochondrial structure in β-cells has not been explored yet. Furthermore, none of the earlier studies have investigated the actions of mitochondrial structure destruction in adult mice. Finally, whether and to what extent they impact secretion stimulated by other agents, including incretins, is less clear. This question is important given that changes in mitochondrial oxidative metabolism (19) and structure contribute to type 2 diabetes (T2D).

Here, we first explored the potential contribution of mitofusins to the effects of diabetic conditions. We next determined whether deletion of Mfn1 and Mfn2 in β-cells in adult mice may impact insulin secretion. Lastly, we aimed to determine whether incretins may rescue or bypass any observed perturbations. We show that mitofusin ablation exerts profound effects on insulin release, glucose homeostasis, and Ca²⁺ dynamics. Remarkably, the deficiencies in insulin secretion are largely corrected by incretin hormones. This suggests a possible approach to ameliorating the consequences of mitochondrial fragmentation with these agonists in some forms of diabetes.

C57BL/6J mice were housed in individually ventilated cages in a pathogen-free facility at 22°C with a 10–14-h light-dark cycle and were fed ad libitum with a standard mouse chow diet (Research Diets, New Brunswick, NJ). All in vivo procedures were approved by the U.K. Home Office, according to the Animals (Scientific Procedures) Act 1986 with local ethical committee (Hammersmith Hospital Campus, London, U.K.) approval under personal project license number PA03F7F07 to I.L.

C57BL/6J male mice bearing Mfn1 (Mfn1^tm2Dcc; JAX stock no. 026401) and Mfn2 (B6.129(Cg)-Mfn2^tm3Dcc/J; JAX stock no. 026525; The Jackson Laboratory, Bar Harbor, ME) alleles (20) with loxP sites flanking exons 4 and 6 were purchased from The Jackson Laboratory and crossed to C57BL/6J transgenic animals carrying an inducible Cre recombinase under Pdx1 promoter control (Pdx1-Cre^ERT2) (21). Mice bearing floxed Mfn alleles but lacking Cre recombinase were used as littermate controls in this study. Mice were genotyped following protocols described by The Jackson Laboratory for each of these strains (see Supplementary Table 1). Recombination was achieved by daily tamoxifen (10 mg/mouse diluted in corn oil; Sigma-Aldrich, Dorset, U.K.) i.p. injections for 5 days at 7–8 weeks of age in both control and β-cell selective Mfn1/Mfn2 deletion knockout (dKO) (βMfn1/2 dKO) groups.

Animals with floxed Clec16a alleles were bred to mice carrying the Pdx1-Cre transgene (Clec16a^Δpanc), as previously described (22). Pdx1-Cre alone mice were used as littermate controls. Pdx1CreER mice were generated as previously described (21).

For measurements of mRNA levels, pancreatic islets from control and βMfn1/2-dKO mice were isolated by collagenase digestion (23). Total RNA from 50 to 100 islets was extracted and underwent quantitative (q)RT-PCR, as previously described (24) (see Supplementary Table 2 for primer details).

Total islet DNA was isolated using Puregene Cell and Tissue Kit (Qiagen, Manchester, U.K.) and was amplified (100 ng) using NADH dehydrogenase I primers (25), also known as complex I (mt9/mt11) for mtDNA and Ndufv1 for nuclear DNA.

Islets were collected and lysed (20 μg), as previously described (24). The antibodies used are summarized in Supplementary Table 3.

Intraperitonal glucose tolerance tests (IPGTTs), intraperitoneal insulin tolerance tests (IPIITTs), oral glucose tolerance tests (OGTTs), and plasma insulin measurements were performed as previously described (24). Plasma proinsulin levels were measured in fasted (16 h) animals using a rat/mouse proinsulin ELISA kit (Mercodia). Plasma β-ketones were measured from fed or fasted (16 h) mice using an Areo 2K device (GlucoMen, Berkshire, U.K.).

Islets were isolated from mice and incubated for 1 h in Krebs-Ringer bicarbonate buffer containing 3 mmol/L glucose, as previously described (24).

Dissociated islets were incubated with 100 nmol/L MitoTracker Green (Thermo Fisher Scientific) in Krebs-Ringer bicarbonate buffer containing 11 mmol/L glucose for 30 min. MitoTracker Green was then washed with Krebs buffer with 11 mmol/L glucose before fluorescence imaging. Experiments with tetramethylrhodamine ethyl ester (TMRE) were performed as previously described (24). Clusters of dissociated islets were transduced for 48 h with an adenovirus encoding the low-Ca²⁺-affinity sensor D4 addressed to the ER, Ad-RIP-D4ER (multiplicity of infection: 100), as previously described (26). Bleaching was corrected as previously described (27). Clusters of dissociated islets were transduced for 24 h with an adenovirus encoding Epac1-camps, as previously described (28).

For each stack, one image at the top, middle, and bottom of the islet was analyzed. After background subtraction, the following parameters were measured for each cell: number of particles, perimeter, circularity, elongation (1/circularity), density, and surface area of each particle (29).

Cytosolic and mitochondrial Ca²⁺ imaging as well as ATP-to-ADP changes in whole islets were performed as previously described (24).

Experiments using the membrane-located zinc sensor ZIMIR (50 μmol/L) (30) or the fluorescent genetically encoded and vesicle-located green marker neuropeptide Y (NPY)-Venus were performed as previously described (31).

Isolated pancreata were fixed and imaged as previously described (24). The antibodies used are summarized in Supplementary Table 3. For examination of apoptosis, TUNEL assay was performed using a DeadEnd Fluorometric TUNEL system kit and DNase I treatment (Promega, Madison, WI), according to the manufacturer's instructions.

Metabolites were quantified using targeted ultrahigh-performance liquid chromatography coupled with triple quadrupole mass spectrometry, as described earlier (32). Lipidomic sample preparation followed the Folch procedure with minor adjustments. Significance was tested by the Student two-tailed t test using GraphPad Prism 8 software.

The oxygen consumption rate (OCR) was measured with XF96 assays (Seahorse Bioscience, Agilent, Santa Clara, CA) using mouse islets (∼10 per well), as previously described (33). Parameters were analyzed as previously described (34).

For conventional electron microscopy, islets were fixed and imaged as previously described (35).

Correlation analyses in an imaged islet were performed as previously described (36).

Processing and differential expression analysis of RNA sequencing data from islets isolated from mice fed a high-fat high-sugar diet (HFHS; D12331, Research Diets) or regular chow (RC) diet (C57Bl/6J, DBA/2J, BALB/cJ, A/J, AKR/J, 129S2/SvPas) was performed as previously described (37) using the Limma package in R. P values were adjusted for multiple comparisons using the Benjamini-Hochberg procedure (38).

Data are expressed as mean ± SD, unless otherwise stated. Significance was tested by the Student two-tailed t test and Mann-Whitney correction or two-way ANOVA with the Šidák multiple comparison test for comparison of more than two groups, using GraphPad Prism 9 software (GraphPad Software, San Diego, CA). P <  0.05 was considered significant. Experiments were not randomized or blinded.

The data sets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request. No applicable resources were generated or analyzed during the current study.

To determine whether the expression of Mfn1 or Mfn2 might be affected under conditions of hyperglycemia mimicking T2D in humans, we interrogated data from a previous report (37) in which RNA sequencing was performed on six mouse strains. BALB/cJ mice showed “antiparallel” changes in Mfn1 and Mfn2 expression in response to maintenance on an HFHS diet for 10 days, and similar changes were obtained in DBA/2J mice at 30 and 90 days (Supplementary Fig. 1A and B).

Efficient deletion of Mfn1 and Mfn2 in the β-cell was achieved in adult mice using the Pdx1-Cre^ERT2 transgene and tamoxifen injection at 7–8 weeks. Possession of this transgene, which does not contain the human growth hormone (hGH) cDNA (21), alone had no effect on glycemic phenotype or cellular composition of pancreatic islets (Supplementary Fig. 2A–C). Deletion of mitofusin genes was confirmed by qRT-PCR (Fig. 1A) and Western (immuno-) blotting (Fig. 1B) analysis, ∼7 weeks posttamoxifen injection. Relative to β-actin, expression of the Mfn1 and Mfn2 transcripts in isolated islets from dKO mice decreased by ∼ 83 and 86% accordingly versus control islets (Fig. 1A), consistent with selective deletion in the β-cell compartment (39). No differences were detected in the expression of other mitochondrial fission and fusion mediator genes such as Opa1, Drp1, and Fis1 in islets (Fig. 1A) or in Mfn1 and Mfn2 in other relevant tissues (Supplementary Fig. 3A). dKO mice were significantly lighter than control animals after 20–21 weeks (Supplementary Fig. 3B).

Glucose tolerance was impaired in dKO mice compared with control littermates at 14 weeks (Fig. 1C and D), and this difference was further exaggerated at 20 weeks (Supplementary Fig. 3 and Fig. 3C). At 14 weeks, βMfn1/2-dKO mice (with a 27 mmol/L glycemia at 15 min) (Fig. 1E and F) showed a dramatically lower insulin excursion upon glucose challenge versus control animals (Fig 1G and H). Following an oral gavage, glucose tolerance was more modestly affected in dKO mice (Fig. 1I and J), while plasma insulin levels in these animals (with a glycemia of 27 mmol/L at 15 min) were indistinguishable from control animals (0 vs. 15 min in dKO) (Fig. 1K and L). Insulin tolerance was unaltered in βMfn1/2-dKO versus control mice (Supplementary Fig. 3D), while proinsulin conversion was impaired (Supplementary Fig. 3E and F). dKO mice displayed significantly elevated plasma glucose (Supplementary Fig. 3G) under both fed and fasted conditions, and β-ketones (ketone bodies) were also elevated in fasted versus control animals (Supplementary Fig. 3H), whereas plasma insulin levels were lower (Supplementary Fig. 3I). Apparent insulin secretion was also impaired after i.p. injection, with a lower glucose in 14- and 20-week-old dKO versus control mice (Supplementary Fig. 4A–D). In contrast, plasma insulin levels were not statistically different between control and dKO animals following an OGTT at either age (Supplementary Fig. 4E–H), although a trend toward lower insulin excursion was evident in dKO mice.

While the mitochondrial network was highly fragmented in dKO cells (Fig. 2A and inset), the number of mitochondria per cell or density was not altered (Fig. 2B). Mitochondrial elongation, perimeter, and surface area were also significantly decreased in βMfn1/2-dKO cells, while circularity was increased (Fig. 2B). Transmission electron microscopy confirmed these changes (Fig. 2C). Cristae structure and organization were also altered in βMfn1/2-dKO cells, with a single crista often running the length of a mitochondrial section. Finally, dKO islets displayed an ∼5% reduction in mtDNA (Fig. 2D).

Pancreatic β-cell mass decreased by 33%, whereas α-cell mass was not affected in dKO mice (Fig. 3A–C). The β-cell–to–α-cell ratio was decreased by 53% (Fig. 3D), in line with an increase in TUNEL-positive β-cells in dKO versus control animals (Fig. 3E and F).

We next sought to exclude the possibility that mitochondrial fragmentation may simply be the consequence of the observed hyperglycemia. Two distinct groups of organelles (both elongated and circular) were apparent in βMfn1/2-dKO cells (Supplementary Fig. 5A and B) 2 weeks after tamoxifen treatment. Neither fed nor fasted glycemia or plasma insulin levels following glucose challenge were different between groups (Supplementary Fig. 5C–E). A trend toward lower β-cell mass and mtDNA was detected in dKO animals (Supplementary Fig. 5F–I).

While Ins2, Ucn3, and Glut2 (Slc2a2) were significantly downregulated, Trpm5 was upregulated in dKO islets (Supplementary Fig. 6). No changes in α- or β-cell disallowed genes (40) were detected. In contrast, genes involved in mitochondrial function, such as Smdt1 and Vdac3, were upregulated in dKO β-cells (Supplementary Fig. 6). Lastly, genes involved in ER stress and mito/autophagy were also affected, with Chop (Ddit3) and p62 being upregulated and Lc3 and Cathepsin L downregulated.

Increased cytosolic Ca²⁺ is a key trigger of insulin exocytosis in response to high glucose (2). dKO mouse islets exhibited a significantly smaller glucose-induced [Ca²⁺]_cyt rise versus control islets (Fig. 4A–C). When the K_ATP channel opener diazoxide and a depolarizing K⁺ concentration were then deployed together to bypass the regulation of these channels by glucose, cytosolic Ca²⁺ increases were not significantly impaired in dKO compared with control animals (Fig. 4B and C). A substantial reduction in mitochondrial free Ca²⁺ concentration ([Ca²⁺]_mito) in response to 17 mmol/L glucose (24) was also observed in dKO islets (Fig. 4D–F). Of note, subsequent hyperpolarization of the plasma membrane with diazoxide caused the expected lowering of mitochondrial [Ca²⁺]_mito in control islets, reflecting the decrease in [Ca²⁺]_cyt (Fig. 4E and F), but was almost without effect on dKO islets.

Glucose-induced increases in Δψ_m were also sharply reduced in dKO versus control mouse islets (Fig. 4G and H). Addition of 2-[2-[4-(trifluoromethoxy)phenyl]hydrazinylidene]-propanedinitrile (FCCP) resulted in a similar collapse in apparent Δψ_m in islets from both genotypes (Fig. 4G). Cytosolic Ca²⁺ oscillations and synchronous Δψ_m depolarization were also largely abolished in response to glucose in dKO cells when measured by intravital imaging in vivo (41). Finally, to assess whether deletion of Mfn1 and Mfn2 may impact glucose-induced increases in mitochondrial ATP synthesis, we performed real-time fluorescence imaging using Perceval (Fig. 4I and J). While control islets responded with a time-dependent rise in the ATP-to-ADP ratio in response to a step increase in glucose from 3 mmol/L to 17 mmol/L, βMfn1/2-dKO β-cells failed to mount any response (Fig. 4J).

Intercellular connectivity is required in the islet for a full insulin secretory response to glucose (42). To assess this, individual Ca²⁺ traces recorded from Cal-520–loaded β-cells in mouse islets (Fig. 4A and B) were subjected to correlation (Pearson r) analysis to map cell-to-cell connectivity (Supplementary Fig. 7A). Following perfusion at 17 mmol/L glucose, βMfn1/2-dKO β-cells tended to display an inferior, although not significantly different, coordinated activity than control cells, as assessed by counting the number of coordinated cell pairs (0.94 vs. 0.90 for control vs. dKO, respectively) (Supplementary Fig. 7C). By contrast, β-cells displayed highly coordinated Ca²⁺ responses upon addition of 20 mmol/L KCl in dKO islets. Similarly, analysis of correlation strength in the same islets revealed significant differences in response to 17 mmol/L glucose between genotypes. In fact, dKO islets had weaker mean β-cell–to–β-cell coordinated activity (0.88 vs. 0.77 for control vs. dKO, respectively; P < 0.05) (Supplementary Fig. 7B and D), indicating that mitofusins affect the strength of connection rather than the number of coordinated β-cell pairs. A tendency toward lower expression of the gap junction gene Cx36/Gjd2 was observed in dKO islets (Supplementary Fig. 7E). β-Cell “hub” and “leader” distributions (43) were also impaired in the dKO group (data not shown; see [41]).

No differences in cytosolic Ca²⁺ responses between genotypes were observed after agonism at the Gq-coupled metabotropic acetylcholine (Ach) receptor (44,45) (Fig. 5A–C). In contrast, measurements of VO₂ revealed that basal, proton leak, and maximal respiratory capacities were significantly impaired in dKO islets (Fig. 5D and E).

While GSIS was markedly impaired in dKO islets (Fig. 6A and Supplementary Table 4), incretins (GLP-1 or GIP), or the GLP1R agonist exendin-4, at a submaximal concentration of 10 mmol/L glucose, led to a significant potentiation in GSIS in both groups. Consequently, insulin secretion in response to 10 mmol/L glucose was no longer different between control and βMfn1/2-dKO islets after incretin addition (Fig. 6A and B). Moreover, under these conditions, forced increases in intracellular cAMP imposed by the addition of forskolin (FSK) or 3-isobutyl-1-methylxanthine (IBMX), which activate adenylate cyclase (AC) and inhibit phosphodiesterase, respectively, eliminated differences in GSIS between the genotypes (Fig. 6B). No differences in insulin secretion were observed between control and dKO islets after depolarization with KCl.

We next explored whether the incretin-mediated improvements in insulin secretion in response to incretins were the result of altered [Ca²⁺]_cyt dynamics. Islets from isolated dKO mice displayed a delayed increase in [Ca²⁺]_cyt in response to 10 mmol/L glucose compared with control islets (Fig. 6C and D). Addition of exendin-4 led to the emergence of oscillatory activity in both groups, and under these conditions, differences between genotypes, as seen in Fig. 4B, were no longer evident (Fig. 6C). Measured at 10 mmol/L glucose, control and dKO islets displayed increases in ER Ca²⁺ in response to exendin-4 (Fig. 6E and F), while the response was exaggerated in the latter group. Neither group displayed significant changes in the ATP-to-ADP ratio in response to exendin-4 (Fig. 6G and H). Analysis of the OCR revealed no significant differences between genotypes at 10 mmol/L glucose in the presence or absence of exendin-4 or FSK (Fig. 6I).

Moreover, mitofusin deletion may lead to a partial activation of “amplification” pathways of GSIS (46) at 3 mmol/L glucose since insulin secretion was enhanced in dKO islets after depolarization of the plasma membrane with KCl in the presence of diazoxide (Fig. 6J). Conversely, no differences between islet genotypes were observed at 17 mmol/L glucose (Fig. 6J).

While glucose-induced β-cell–β-cell connectivity, as assessed by monitoring Ca²⁺ dynamics (Fig. 6C), was markedly impaired in dKO islets (Fig. 7A and Supplementary Fig 7), these differences were largely abolished in the presence of exendin-4 (Fig. 8B–D).

To explore the actions of mitochondrial disruption on incretin signaling, we next used a pharmacological approach. GSIS was more strongly enhanced in dKO versus control islets by IBMX, FSK, or the protein kinase A (PKA) inhibitor H89 alone (Fig. 8A and Supplementary Table 4). Selective activation of EPAC also tended to lead to a larger increase in insulin secretion in dKO than in control islets, and this difference became significant when PKA was inhibited with H89 (Fig. 8B).

Glucose-dependent increases in cytosolic cAMP, assessed using the Epac-camps sensor, were also markedly amplified in dKO versus control cells (Fig. 8C and D). This difference persisted in the presence of IBMX and FSK, added separately or alone (Fig. 8C and E). No changes in the expression of Epac, Adcy, or Prkar (PKA) subunits were apparent between control and dKO islets (Fig. 8F).

To determine whether incretins may reverse defective insulin secretion in an alternative model of mitochondrial dysfunction, we examined mice lacking the mitophagy regulator Clec16a selectively in the pancreatic islet (Clec16a^Δpanc) (22). GSIS was sharply inhibited in null versus Pdx1-Cre control mice, and these differences between genotype were largely corrected in by the addition of exendin-4 (Supplementary Fig. 8A). Correspondingly, whereas the difference between Clec16a^Δpanc and control mice was significant for IPGTTs, there was no such (significant) difference for the OGTTs at 15 min, in line with the findings above for βMfn1/2-dKO mice (Supplementary Fig. 8A–C).

To determine whether the markedly weaker stimulation of insulin secretion in dKO islets may reflect failed recruitment of secretory granules into a readily releasable or morphologically docked pool beneath the plasma membrane, we next deployed total internal reflection fluorescence microscopy in dissociated β-cells. By overexpressing NPY-Venus, the number of insulin granules was significantly higher in close proximity with the plasma membrane in dKO cells after treatment with 20 mmol/L KCl (Supplementary Fig. 9A and B). However, when we then used ZIMIR (30) in response to depolarization as a surrogate for insulin secretion, release events were fewer in number and smaller in dKO (Supplementary Fig. 9C–E).

We applied an -omics approach to study metabolite and lipid changes in peripheral plasma samples from control and dKO mice (Supplementary Fig. 10). Of 29 metabolites, the levels of five metabolic species (shown in red) were significantly altered in βMfn1/2-dKO animals (Supplementary Fig. 10A). In the lipidomics analysis, the majority of lipid classes displayed a remarkably homogeneous downward trend in dKO samples (Supplementary Fig. 10B).

The key goal of the current study was to determine the role of mitofusins in controlling mitochondrial dynamics and hence glucose- and incretin-stimulated insulin secretion in the β-cell. Our strategy involved deleting both mitofusin isoforms since the expression of Mfn1 and Mfn2 is similar in the β-cell (47), suggestive of partial functional redundancy (48). Our measurements of Mfn1 and Mfn2 expression in mouse models of T2D nonetheless revealed changes in the expression of these genes, which may contribute to the disease.

Importantly, we show that Mfn1 and Mfn2 are critical regulators of the mitochondrial network in β-cells and consequently of insulin secretion in vitro and in vivo (see also [41]) (Supplementary Fig. 11A and B). These findings are in line with earlier studies, albeit involving the deletion of genes other than the mitofusins (13–18). Additionally, we show that changes in Mfn1 and Mfn2 expression occur in models of diabetes, and hence, their forced changes, as achieved in our study, may have relevance for the pathoetiology of β-cell failure in T2D and metabolic changes consistent with insulin deficiency. These include higher levels of bile acids as previously described in rodent models of type 1 diabetes (T1D) and T2D and in humans (49,50), elevated leucine and isoleucine, as observed in human T1D (51), and an altered triglyceride profile (52). Finally, these metabolomic/lipidomic data provide further support for the expected actions of mitofusin deletion via altered β-cell function, with changes that are somewhat more in line with metabolomic changes in human T1D (and models thereof) than T2D (53). Indeed, dKO mice gain less weight than controls as they show the classic symptoms of diabetes (54,55). This is likely to be the result of metabolic dyshomeostasis in the face of lowered circulating insulin levels, leading to impaired fat storage, loss of liver and muscle glycogen, and eventually, loss of muscle mass (i.e., the cardinal symptoms of T1D and of advanced insulin-requiring T2D in humans).

Of note, none of the earlier reports investigating the effects of mitochondrial disruption in the β-cell explored the effects on incretin-stimulated secretion. Suggesting a differential effect on glucose- versus incretin-stimulated secretion we show here; firstly, that insulin secretion and glucose excursion were less markedly affected by mitofusin knockout during OGTTs, where an incretin effect is preserved (56), than during IPGTTs. Correspondingly, insulin secretion stimulated by incretins was largely preserved in dKO cells, in contrast to the ablation of glucose-stimulated secretion (Supplementary Fig. 11C and D). Strikingly, mitofusin deletion also enhanced incretin-stimulated cytosolic cAMP increases. That this effect was preserved in the face of phosphodiesterase inhibition (IBMX) and AC activation was surprising but may reflect an increase in total AC activity or distribution in dKO cells.

While PKA suppression is considered to be either neutral or inhibitory toward GSIS in wild-type β-cells (57–59), our data show a rather striking increase in insulin secretion in the presence of H89 in islets from mice of either genotype. While unexpected, and in contrast with those of others that support a role for PKA downstream of cAMP in the β-cell, Bryan and colleagues (57) provide some evidence for the stimulation of GSIS by H89 under certain conditions. Nevertheless, several studies have stressed the importance of both PKA-dependent and PKA-independent effects of increased [cAMP]_i on GSIS from islets (60). Thus, PKA-independent exocytosis occurs through interactions between Epac-2/cAMP- guanine-nucleotide-exchange factor II (61,62), Rab3A, and Rim2 (proteins involved in vesicle trafficking [57,58, 63] and fusion) (64). On the other hand, GLUT2, Kir6.2, and SUR1, and α-SNAP (a vesicle-associated protein) have been reported to be phosphorylated by PKA (58). Here, we show that the effect of mitofusin deletion on GSIS is preserved when PKA is inhibited by H89 and even potentiated by EPAC-activation (Supplementary Fig. 11C and D). These changes appear to be exerted at the posttranscriptional level, since we observed no changes in levels of mRNAs encoding the relevant β-cell isoforms of Epac. Whether there are changes in the level or the corresponding proteins including EPAC, their subcellular localization or interaction with upstream regulators or downstream effectors, remains to be explored. Finally, the latter findings could indicate that an intact mitochondrial reticulum restricts signaling by EPAC through a mechanism that is inhibited by PKA. Future studies, using additional or alternative PKA inhibitors (65), will be needed to explore these possibilities.

Possibly contributing to these differences in the effects on responses to glucose versus incretin, exendin-4 treatment led to greater Ca²⁺ accumulation in the ER in dKO cells. By enhancing Ca²⁺ cycling across the ER membrane, this could conceivably drive larger local increases in cytosolic Ca²⁺, which, in turn, may influence plasma membrane potential, trigger Ca²⁺ influx via VDCCs, and hence, stimulate insulin release (66).

We also demonstrate that preserved mitochondrial ultrastructure is critical for normal β-cell–β-cell connectivity, itself required for normal insulin secretion (41,67). The mechanisms underlying impaired connectivity in the absence of mitofusins are unclear but may involve altered Cx36/Gjd2 expression, phosphorylation, or activity impacting gap junctions (42).

In summary, we show that acute treatment with incretins, commonly used as treatments for T2D and obesity (56), largely reverses the deficiencies in insulin secretion that follow mitochondrial disruption. Future studies will be needed to address the relevance of these findings to human β-cells and to the action of incretins in clinical settings.

Acknowledgments. The authors thank Stephen M. Rothery, from the Facility for Imaging by Light Microscopy (FILM) at Imperial College London, for support with confocal and widefield microscopy image recording and analysis. The authors thank Professor Julia Gorelik and Sasha Judina (Imperial College) for providing the Epac1-camps sensor, and Aida Di Gregorio from the National Heart and Lung Institute (Imperial College) for genotyping the mice.

Funding. G.A.R. was supported by a Wellcome Trust Senior Investigator Award (098424AIA) and Wellcome Trust Investigator Award (212625/Z/18/Z), Medical Research Council Programme grants (MR/R022259/1, MR/J0003042/1, MR/L020149/1), an Experimental Challenge Grant (DIVA, MR/L02036X/1), a Medical Research Council grant (MR/N00275X/1), and Diabetes UK grants (BDA/11/0004210, BDA/15/0005275, BDA16/0005485). I.L. was supported by a Diabetes UKD project grant (16/0005485). This project has received funding from the European Commission Innovative Medicines Initiative 2 Joint Undertaking, under grant agreement no. 115881 (RHAPSODY). This Joint Undertaking receives support from the European Union’s Horizon 2020 Research and Innovation Programme. This work is supported by the Swiss State Secretariat for Education, Research and Innovation (SERI), under contract no. 16.0097. A.T. was supported by Medical Research Council project grant MR/R010676/1. Intravital imaging was performed using resources and/or funding provided by National Institutes of Health grants R03 DK115990 (to A.K.L.), Human Islet Research Network UC4 DK104162 (to A.K.L., RRID:SCR_014393). BJ acknowledges support from the Academy of Medical Sciences, Society for Endocrinology, The British Society for Neuroendocrinology, the European Federation for the Study of Diabetes, an Engineering and Physical Sciences Research Council capital award, and the Medical Research Council (MR/R010676/1). S.A.S. was supported by the JDRF (CDA-2016-189, SRA-2018-539, COE-2019-861), the National Institutes of Health (R01 DK108921, U01 DK127747), and the U.S. Department of Veterans Affairs (I01 BX004444).

Duality of Interest. This Joint Undertaking receives support from the European Federation of Pharmaceutical Industries and Associations. G.A.R. has received grant funding and consultancy fees from Les Laboratoires Servier and Sun Pharmaceuticals. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. E.G. performed experiments and analyzed data. E.G. supported the completion of confocal and widefield microscopy and analysis. E.G. contributed to designing the study and writing the manuscript C.M., M.M., and A.K.L. were responsible for the in vivo intravital Ca²⁺ imaging in mice presented in the bioRxiv paper [41]. P.C. contributed to the analysis and manipulation of the in vivo intravital Ca²⁺ measurements as well as the preparation and imaging of total internal reflection fluorescence samples. E.A. and L.L.N. performed the oral gavage in live animals. A.T. performed the electron microscopy sample processing and data analysis. F.Y.S.W. and Y.A. generated and performed Monte Carlo-based signal binarization. T.S. contributed to the generation of the MATLAB script used for connectivity analysis. A.W. and C.L.-Q. contributed to the metabolomics analysis. B.J. assisted with the cAMP assays. Y.X. and G.G. performed studies with the Pdx1CreER mice. N.A. assisted with Seahorse experiment protocols. C.C.-G., C.M., and M.I. were responsible for the RNA sequencing data analysis. I.L. and T.A.R. were responsible for the maintenance of mouse colonies and final approval of the version to be published. S.A.S. performed studies with Clec16a mice. T.A.R. was involved in the design of the floxed Mfn alleles. G.A.R. designed the study and wrote the manuscript with input and final approval of the version to be published from all authors. G.A.R. is the guarantor of this work and, as such, had full access to all 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 an oral or poster presentation at the 81st Scientific Sessions of the American Diabetes Association, virtual meeting, 25–29 June 2021; the Australasian Diabetes Congress 2021, virtual event, 11–13 August 2021; the 80th Scientific Sessions of the American Diabetes Association, virtual meeting, 12–16 June 2020; Diabetes UK Professional Conference 2019, Liverpool, U.K., 6–8 March 2019; Gordon Research Conference, New London, NH, 19–22 March 2019; Rhapsody Consortium; and the 54th Annual Meeting of the European Association for the Study of Diabetes, Berlin, Germany, 1–5 October 2018. A non–peer-reviewed version of this article was published on the bioRxiv preprint server (https://doi.org/10.1101/2020.04.22.055384) on 24 April 2020.