Pancreatic beta cell selective deletion of mitofusins 1 and 2 (Mfn1 and Mfn2) disrupts mitochondrial architecture and abrogates glucose-stimulated insulin secretion in vivo

1Section of Cell Biology and Functional Genomics, Division of Diabetes, Endocrinology and Metabolism, Department of Medicine, Imperial College London, London, UK 2Center for Diabetes and Metabolic Diseases, Indiana University School of Medicine, Indianapolis, IN, USA 3Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore 4Loughborough University, Centre of Innovative and Collaborative Construction Engineering, Leicestershire, UK 5FILM, Imperial College London, London, UK 6National Heart and Lung Institute, Imperial Centre for Translational and Experimental Medicine, Imperial College London, London, UK


Introduction
Mitochondria are often referred to as the powerhouses, or more recently, as the "chief executive organelles" (CEO) of the cell, generating most of the energy required to sustain normal function. Mitochondria are responsible for nicotinamide adenine dinucleotide (NAD) and hydrogen (NADH) production via the Krebs cycle, oxidative phosphorylation, adenosine triphosphate (ATP) synthesis, fatty acid oxidation and gluconeogenesis, and mitochondrial DNA (mtDNA) distribution. They also contribute to the regulation of apoptosis and their own turnover via mitophagy [1].
Mitochondrial oxidative metabolism plays a pivotal role in the response of pancreatic beta cells to stimulation by glucose and other nutrients [2][3][4]. Thus, as concentrations of sugar increase in the blood, enhanced glycolytic flux and oxidative metabolism in these cells lead to an increase in ATP synthesis, initiating a cascade of events which involve the closure of ATP-sensitive K + (KATP) channels [5], plasma membrane depolarisation and the influx of Ca 2+ via voltage-gated Ca 2+ channels (VDCC). The latter mechanism, along with other, less well defined amplifying signals [6], drive the biphasic release of insulin [4].
The possibility that changes in mitochondrial function in these cells may contribute to declining insulin secretion and to type 2 diabetes (T2D) has been the subject of extensive investigation [7]. Reduced glucose-stimulated insulin secretion (GSIS) in beta cells, alongside altered mitochondrial function, dynamics and morphology, were observed in diabetic models [8][9][10].
Besides ATP, mitochondria are also one of the main producers of reactive oxygen species (ROS) in the cell, inducing oxidative stress and tissue damage [11].
Mitochondrial dysfunction in type 2 diabetic or obese patients suffering from hyperglycaemia was shown to be directly linked with over-production of ROS, lowered ATP levels and mitochondrial content as well as the development of insulin resistance [12]. Additionally, several mtDNA variations in human populations were associated with increased or decreased risk of T2D [13], while in animal models, alterations in beta cell mtDNA led to reduced insulin secretion, hyperglycemia and beta cell death [14].
The multifaceted roles of mitochondria in the cell are associated with an equally variable morphology. Under normal physiological conditions, these organelles repetitively undergo fusion and fission cycles which are essential for their quality control and adaptation to energetic demands [15]. Thus, highly inter-connected mitochondrial networks allow communication and interchange of contents between mitochondrial compartments, as well as with other organelles such as the endoplasmic reticulum (ER) [16]. These exist interchangeably with more fragmented structures, with the balance between the two influenced by external stimuli and metabolic demands [17]. Thus, in many cell types including INS1, a pro-fused state is often observed during starvation or acute oxidative stress where energy production is required for cytoprotection and resistance to apoptosis. The opposite pertains when cells are exposed to increased nutrient supply such as in obesity or T2D [18]. On the other hand, exposure to a nutrient overload can stimulate mitochondrial fission and uncoupled respiration [18], but is critical for the elimination of damaged mitochondria by mitophagy [19].
Over the past two decades, considerable light has been shed on the molecules and mechanisms that control mitochondrial dynamics. The mitofusins MFN1 and MFN2 are key mediators of mitochondrial outer membrane (OMM) fusion and OPA1 of the inner mitochondrial membrane (IMM) fusion, while DRP1 is responsible for mitochondrial fission [20].
In pancreatic beta cells, defects in mitochondrial structure, accumulation of damaged or depolarised organelles are associated with oxidative stress and the subsequent development of diabetes in patients and animal models [15]. Changes in mitochondrial fusion and fission dynamics are observed in the pancreatic beta cell in animal models of diabetes [7,21,22], and patients with T2D and obesity exhibited smaller and swollen mitochondria in pancreatic tissue samples [23]. The latter, was suggested to be a consequence of hyperglycemia [24], followed by increased ROS production demonstrating that mitochondrial dynamics can be master regulators of mitochondrial activity [25].
Suggesting that changes in the expression of mitochondrial fusion regulators may also be altered in diabetes in extra-pancreatic tissues, insulin resistant and obese patients show reduced MFN2 expression in skeletal muscle [26]. Possibly resulting from altered expression of the peroxisome proliferator-activated receptor-gamma coactivator (PGC)-1alpha and the estrogen-related receptor alpha (ERRα) [27,28].
This lead to increases in ROS production and impaired mitochondrial respiration in both skeletal muscle and liver tissues through the JNK pathway [27][28][29]. Whether these changes are the cause, or the consequence, of impaired mitochondrial dynamics, remains unclear, however [30][31][32].
To address this question in the context of the pancreatic beta cell, we explore the impact of targeted deletion of Mfn1 and Mfn2, specifically in mouse beta cells in adult animals. We show that this exerts profound effects on beta cell mass, insulin secretion and glucose homeostasis.  Table   2 for primer details).

IPGTT test and measurement of insulin secretion in vivo
To investigate glucose tolerance, male mice (aged 14 weeks) were fasted overnight and injected with glucose (20% w/v, 1g/kg body weight) IP. Glucose was measured in tail vein blood at time points as indicated using an ACCU-CHECK Aviva glucometer (Roche). For in vivo insulin secretion measurements, fasted male mice were administered glucose (20% w/v, 3g/kg body weight) IP, and plasma insulin was measured using an ultra-sensitive mouse insulin enzyme-linked immunosorbent assay (ELISA) kit (CrystalChem).

Mitochondrial shape analysis
To determine morphological characteristics of mitochondria, confocal stacks were analysed with ImageJ using an in-house macro (available upon request). Briefly, for each stack, one image at the top, middle and bottom of the islet was analysed. After background subtraction, the following parameters were measured for each cell: number of particles, perimeter and circularity of each particle and elongation (1/circularity) was calculated [37]. The average perimeter, circularity and elongation of particles was then calculated for each cell.
Whole-islet fluorescence imaging Ca 2+ imaging of whole islets was performed after infection with adenovirus encoding the mitochondrially-targeted probe, Pericam Fluorescent traces were calibrated to calculate fluorescence spikes fold change above the baseline on three cells per condition using ImageJ. Then the AUC was determined, with a threshold at 0.15 to subtract background noise.

Connectivity analysis
Pearson-R correlations. Significantly correlated beta cell pairs were measured on islets incubated with Cal-520 as described previously [39,40]. Correlation coefficient (R) and heatmaps were generated using an in-house MATLAB script (available upon request).
Monte Carlo-based signal binarisation and data shuffling for identification of highly connected cells Data were analysed using approaches similar to those previously described [39,40] with modifications as below. Ca 2+ signals were denoised by subjecting the signal the Huang-Hilbert type (HHT) empirical mode decomposition (EMD). The signals were decomposed into their intrinsic mode functions (IMFs) in MATLAB [41]. The residual and the first IMF with the highfrequency components were then rejected to remove random noise.
The Hilbert-Huang Transform was then performed to retrieve the instantaneous frequencies [42][43][44] of the other IMFs to reconstruct the new signal using where aj(t) = amplitude, ωj(t) = frequency of the ith IMF component to retrieve a baseline trend and to account for any photobleaching or movement artefacts. A 20% threshold was imposed to minimise false positives from any residual fluctuations in baseline fluorescence.
Cells with deflection above the de-trended baseline were represented as '1' and inactivity represented as '0', thus binarising the signal at each time point. The coactivity of every cell pair was then measured as: where Tij = total coactivity time, Ti and Tj = total activity time for two cells.
The significance at p<0.001 of each coactivity measured against chance was assessed by subjecting the activity events of each cell to a Monte Carlo simulation [45,46] with 10,000 iterations.
Synchronised Ca 2+ -spiking behaviour was assessed by calculating the percentage of coactivity using the binarised cell activity dataset. A topographic representation of the connectivity was plotted in MATLAB with the edge colours representing the strength of the coactivity between any two cells.
An 80% threshold was imposed to determine the probability of the data, which was then plotted as a function of the number of connections for each cell to determine if the dataset obeyed a power-law relationship [47].
Statistical Analysis Data are expressed as mean ± SEM. Significance was tested by Student's two-tailed t-test, one or two-way ANOVA with Sidak's or Bonferroni multiple comparison test for comparison of more than two groups, using GraphPad Prism 8 software. p<0.05 was considered significant.
Body weight did not differ between groups over the first seven weeks post-tamoxifen injection (Fig. 1B). Following this period, the mean body weight of βMfn1/2-KO mice dropped significantly at 21-22 weeks (p<0.05), representing a 13% weight loss compared to the WT mice. We next explored the consequences of Mfn1/2 deletion with regards to GSIS from isolated βMfn1/2-KO islets at 14 weeks of age. With respect to WT controls, the latter displayed a sharply significant blunting in the secretory response to low and high glucose concentrations but also to depolarisation with KCl with near complete elimination of insulin secretion in response to stimulation (p<0.05; performed on βMfn1/2-KO and WT mice at 14 weeks of age ( Fig. 2A). The challenge revealed an impaired glucose tolerance in βMfn1/2-KO mice compared to their control littermates with levels of glucose being higher at all time points following glucose injection ( Fig. 2A-B; p<0.05, p<0.01). Insulin concentrations were also measured following a 3g/kg body weight glucose IP injection (Fig. 2C-D) during which, plasma was sampled at 0, 5, 15 and 30 min (Fig. 2E). Basal glucose levels were modestly, but significantly higher in KO mice during all glucose challenges. βMfn1/2-KO mice showed dramatic lower insulin levels upon glucose challenge suggesting an insulinsecretory deficiency ( Fig. 2E-F, p<0.05, p<0.001, p<0.0001). shape as observed in the enlarged confocal images ( Fig. 3B; p<0.0001). Last, mitochondrial structure was evaluated in isolated islets by TEM and revealed highly fragmented mitochondria in KO mice compared to the WT group (Fig. 3C). Cristae structure and organisation was also highly altered in βMfn1/2-KO cells. More precisely, cristae invaginations were completely absent, as observed in the enlarged panels (Fig. 3C).

Deletion of Mfn1/2 alters mitochondrial morphology in beta cells
Mitofusins are essential in maintaining normal glucose-stimulated Ca 2+ uptake and mitochondrial membrane potential in beta cells Since cytosolic Ca 2+ is a major inducer of insulin exocytosis, Ca 2+ dynamics in whole islets ( Fig. 4A-B). The KATP channel opener diazoxide and a depolarising K + concentration (20 mmol/l KCl) were then deployed together to bypass glucose regulation. Under these conditions, cytosolic Ca 2+ increases were also impaired compared to WT animals (AUC, p<0.01; Fig. 4A-B).
Since Ca 2+ entry into the mitochondrial matrix is likely to be important for the stimulation of oxidative metabolism [48], we then determined the impact of deleting ]mito in intact islets, where a substantial difference in response to 17 mmol/l glucose was observed between the two groups ( Fig. 4C-D). Additionally, glucoseinduced increases in mitochondrial membrane potential (Δψm), assessed with TMRE on 14 weeks dissociated beta cells, were abrogated in KO mice which were unable to sequester TMRE (AUC, p<0.01; Fig. 4E-F). FCCP was used to induce depolarisation and completely eliminate Δψm.

Beta cell-beta cell connectivity is impaired by Mfn1/2 ablation Intercellular
connectivity is required in the islet for a full insulin secretory response to glucose [10,40]. To assess this, individual Ca 2+ traces recorded from 14-weeks-Cal-520-loaded beta-cells in mouse islets (Fig. 4A) were subjected to correlation (Pearson R) analysis to map cell-cell connectivity. Following perfusion at 17 mmol/l glucose, βMfn1/2-KO beta cells displayed an inferior, but not significantly different coordinated activity than WT cells (Fig. 5A), as assessed by counting the number of coordinated cell pairs ( Fig. 5C;  than the number of coordinated beta cell pairs. We also explored beta cell connectivity using data binarisation and network theory to determine whether a hierarchy existed in the degree to which individual beta cells were coupled across the islet [39,40]. Clear adherence to a power law distribution of connected beta cells was apparent in both islet groups in the elevated glucose condition (~10% of "hub" cells hosted >20% of all connections to other beta cells; R 2 =0.21 for WT and R 2 =0.45 for KO islets) (Supp. Fig. 1). Thus, whilst lowering overall connectivity as assessed by either Pearson or Monte Carlo analysis, a similar proportion of cells served as hubs in each case.
Mitofusin deletion leads to beta cell loss Immunohistochemical analysis in mice sacrificed at 14 weeks showed that deletion of Mfn1 and Mfn2 caused a gradual loss (~33%) of pancreatic beta (insulin-positive) cells in the KO group (p<0.05; Fig. 6A-B). Alpha (glucagon-positive) cell surface was not affected by loss of mitofusin genes (Fig. 6C). However, Mfn1 and Mfn2 loss was associated with a ~53% reduction in beta cell-alpha cell ratio (p<0.05; Fig. 6D).

Glucose-induced cytosolic Ca 2+ and Δψm changes are impaired in beta cells in
vivo Two-photon imaging was adopted to allow visualisation of the intact pancreas, exposed through an abdominal incision on anaesthetised mice [49]. Islets previously infected in vivo with GCaMP6s and co-stained with TMRM immediately prior to data capture, were imaged for 18 min during which, cytosolic Ca 2+ oscillations and mitochondrial membrane depolarisation were recorded post IP injection of glucose in WT (Fig. 7A) and KO (Fig. 7B)  islets.

Discussion
Mitochondrial dynamics contribute to the maintenance of a metabolically-efficient mitochondrial population, with an impaired balance between fusion and fission impacting mitochondrial morphology and functionality. To date, dynamic and bioenergetic mitochondrial defects have been explored in conditional KO mouse models where different organs such as the liver, skeletal muscle, adipocyte, heart, nervous system, placenta and optic nerve were studied [50]. production and apoptosis [56] suggesting that a balance between fission and fusion is critical to avoid pathological changes. In line with these data, mice deficient for Opa1 in the beta cell develop hyperglycemia, and show defects in ETC complex IV which lead to compromised glucose-stimulated ATP production, oxygen (O2) consumption, Ca 2+ dynamics, and insulin secretion [57].
In this study, we report that disturbance of the balance between fusion and fission leads to a profound disruption of beta cell function. Thus, using a novel mouse model to achieve highly efficient and selective ablation of Mfn1 and Mfn2 in pancreatic beta cells, we reveal that an intact mitochondrial network is essential to maintain normal insulin secretion and circulating glucose levels. Mitochondrial adaptation to altered physiological conditions relies on the regulation of mitochondrial morphology, especially at the level of cristae compartment [58].
Changes in cristae number and shape delineate the respiratory capacity as well as cell viability [58]. Hence, altered assembly of IMM invaginations or even, loss of cristae may have a functional relevance to our animal model phenotype. Since cristae host the respiratory chain components, it is practical to hypothesise that more cristae translate into more respiratory capacity. An impaired cristae assembly could be synonymous with defective organisation of ETC complexes and lower ATP production, explaining the drop in insulin secretion observed in vivo and in vitro. where highly depolarised mitochondria were depicted in KO beta cells. On the other hand, WT beta cell mitochondria responded to increasing glucose concentrations with a corresponding increase in Δψm, which correlated with an increase in insulin secretion [56]. It would also be interesting to investigate how ER homeostasis and Ca 2+ storage and release are affected in KO cells since cytosolic and mitochondrial Ca 2+ oscillations were deleteriously affected.

Under normal conditions, glucose-induced increases in cytosolic Ca
Interestingly, previous studies [62] have shown that inactivation of Mfn1 and Mfn2 from the mouse heart leads to deficiencies in cardiac performance. Loss of beta cell identity is now thought to be a key element driving impaired beta cell function in T2D [4,63] and may underlie impaired insulin secretion from remaining beta cells in type 1 diabetes [64]. As glucose-induced insulin secretion was almost eliminated in vivo and in vitro after Mfn1/2 deletion in the beta cell, despite only partial loss of beta cells, our data support a similar change as contributing to functional alterations in the current model. Future studies, including measurements of the expression of key beta cell identity (and "disallowed") [65] genes may be informative on this point.
Another finding of our study was that the strength of coordinated beta cell responses to 17 mmol/l glucose, using Pearson R analysis, was affected in βMfn1/2-KO mice while no changes were identified in the number of glucose-responsive cells. This was also not associated with a loss of hierarchical behaviour and the loss of clearly identifiable "hub" cells following Monte Carlo simulation. Thus, we demonstrate that mitochondrial ultra-structure is important for normal beta-beta cell connections which are, in turn, essential for coordinated Ca 2+ influx into beta cells and ultimately, efficient and oscillatory insulin secretion [39].
Immunofluorescence analysis showed reduction in beta cell volume (the ratio of beta cell over total pancreas area) in KO islets describing the occurrence of beta cell loss and dysfunction in T2D [66]. This was additionally confirmed by a reduced yield and smaller size of islets isolated from βMfn1/2-KO mice. The underlying mechanisms for the reduction of beta cell mass in βMfn1/2-KO animals seems likely to be a dramatically increased rate of beta cell apoptosis [67] though we do not exclude altered beta cell replication or neogenesis [66] as contributors. Previous studies have raised a number of potential mechanisms by which mitochondrial fragmentation can affect apoptosis [68,69]. Fragmentation occurs early in the cell death pathway [70] and is thought to contribute to cytochrome c production and release into the cytosol, and decrease in O2 consumption, thereby enhancing oxidative stress and ROS production [71]. Future measurements of O2 consumption rates and ROS production will be needed to explore these possibilities.
Other cellular signals have also been reported to induce a fragmented mitochondrial network by accelerating fission and inhibiting fusion [72]. These include extracellular signal-regulated protein kinases 1 and 2 (ERK1/2), known to inhibit fusion by phosphorylating DRP1 in cultured neuronal or tumour cells [72][73][74] In conclusion, we provide evidence that an altered balance of mitochondrial fusion and fission has a drastic impact on beta cell function. Our findings establish an important role for beta cell Mfn1 and Mfn2 in regulating plasma glucose levels, GSIS, islet Ca 2+ oscillations and beta cell mass. Further investigation on the role of mitochondrial dynamics in insulin secretion, as well as, how changes in the expression of mitochondrial dynamics proteins could contribute to T2D would appear to be justified.

G.A.R. received grant funding from Les Laboratoires Servier and was a consultant
Sun Pharmaceuticals.
Data availability All data generated or analysed during this study are included in the published article (and its supplementary information files). No applicable resources were generated or analysed during the current study.           Table 1 Mfn1 F-TGGTAATCTTTAGCGGTGCTC

R-CATCCATGGCGAACTGGTG
List of primers used for qRT-PCR amplification (F: forward, R: reverse).