Mitochondrial metabolism and dynamics in pancreatic beta cell glucose sensing

Type 2 diabetes (T2D) affects almost 1 in 10 adults worldwide, with the total number of individuals affected by the disease predicted to increase to >700m by 2045 (IDF.org). As revealed in the 1990s by the U.K. prospective diabetes study [1], a fall in insulin sensitivity during pre-diabetes is initially compensated by increased insulin production. When the pancreas fails, hyperglycemia and frank diabetes ensue. Other data suggest that functional beta cell loss occurs early in the disease process [2]. Further highlighting the critical role of beta cell deficiency, the majority of genetic variants associated with disease risk identified in genome-wide association studies affect insulin production rather than action [3,4]. Although decreases in beta cell mass may occur in T2D (as reported in cross-sectional studies [5]), other work [6] suggests that these changes are relatively small (but might nevertheless be consequential if they affect subsets of beta cells with critical roles) [7–9]. Defective glucose-stimulated insulin secretion [10], and possible loss of cellular identity [11,12], are now thought to be the more important drivers of disease development and progression.

Here, we discuss firstly the evidence that mitochondrial oxidative metabolism is essential for beta cell glucose sensing, explaining how these organelles integrate both nutritional and receptor-derived signals. We next explore how mitochondrial structure and dynamics contribute to both the activity of individual beta cells and, potentially, to the ensemble behavior of these cells within the islet micro-organ. Finally, we discuss the possibility that mitochondrial dynamics may represent a therapeutic target in T2D.

Studies by Coore and Randle [13] in Cambridge and by Grodsky et al. [14] in San Francisco, established the ‘substrate site hypothesis' for glucose control of insulin secretion, in which metabolic signaling by the sugar plays the key role. Amongst the first evidence supporting this theory were the observations that the dose response for glucose metabolism closely matched that of insulin secretion. Furthermore, the stimulatory effects of glucose were mimicked by metabolizable (mannose), but not by non-metabolizable (galactose) sugars. It was later concluded that the rapid equilibration of glucose across the beta cell plasma membrane (catalyzed by glucose transporters GLUT2 [SLC2A2] in rodents and GLUT1–3 [SLC2A1–SLC2A3] in human beta cells) [15] is followed by glucose phosphorylation: both glucose phosphorylation and insulin secretion are inhibited by mannoheptulose [16]. Glucose phosphorylation, catalyzed by glucokinase (see below), thus represents the flux generating and ‘rate limiting' step (i.e. that with the highest control strength) [17] for glucose metabolism, and hence in the triggering of insulin secretion [18,19]. Metabolism through glycolysis and the citrate cycle, and increases in cytosolic ATP/ADP ratio, then impact plasma membrane potential and Ca²⁺ influx to trigger the fusion of secretory granules at the plasma membrane, and ultimately insulin release [20,21] (Figure 1). Additional, and less well-defined pathways (dubbed ‘amplifying') [22,23] further potentiate the effects of glucose and the actions of Ca²⁺ on insulin exocytosis (see later sections). Of note, and as described by Ashcroft et al. [24], reductive metabolism of glucose via the pentose phosphate shunt does not appear to be involved in the stimulation of insulin secretion. However, more recent studies [25], using time resolved metabolomics, have questioned this conclusion.

Hormones and neurotransmitters, usually acting through G protein-coupled receptors and including the ‘incretins' released from the gut during food transit [26], further modulate insulin release via cAMP generation and activation of beta-arrestins [27] (see later sections).

Mitochondria generate 36 ATP per glucose molecule oxidized versus the two generated by anaerobic glycolysis, suggesting an important contribution of mitochondrial bioenergetics to the triggering of secretion by glucose, as indicated by early observations of Randle, Grodsky and co-workers. Supporting such a role are the inhibitory effects on secretion of mitochondrial uncouplers, as observed by Coore and Randle [13], and subsequent findings from Panten et al. [28], and Hutton et al. [29], showing that extracellular fuels metabolized only (or primarily) by mitochondria e.g. leucine and ketoisocaproate, are efficient stimulators of secretion [30]. Studies by Ashcroft et al. [31] and by Cook and Hales [32], later revealed the critical importance of closure of ATP-sensitive K⁺ (K_ATP) channels (Figure 1). As such, the latter findings described a molecular target both for alterations in ATP/ADP ratio and for the sulphonylurea class of anti-hyperglycemic drugs which bind to the SUR subunit of K_ATP [13,33]. Whilst other targets for ATP/ADP have been proposed, including AMP-activated protein kinase (AMPK) [34], current evidence suggests that the role of the latter enzyme in the control of insulin secretion is complex, and time-dependent [35]. Interestingly, the upstream AMPK kinase, liver kinase B1 (LKB1/STK11), is also required for normal mitochondrial function in beta cells [36]. However, and in contrast with the effects of AMPK inactivation, deletion of LKB1 lowers the requirement of glucose-stimulated insulin secretion for mitochondrial function, through as-yet poorly understood mechanisms that involve the activation of ‘amplifying' pathways [36].

Critical roles for mitochondrial metabolism in glucose-regulated insulin secretion were also inferred from human genetics. Thus, a mutation in the transfer RNA tRNAleu, required for the efficient translation of mitochondrial DNA (mtDNA), leads to maternally inherited diabetes and deafness [37,38]. Likewise, variants in the TFB1M gene, encoding another mitochondrial transcription factor, are associated with increased risk of T2D [39]. Corroborating these findings, deletion of the mitochondrial transcription factor and mtDNA stabilizing factor TFAM caused diabetes in mice [40].

Further studies revealed that the beta cell has an unusual enzymatic configuration which strongly favors oxidative versus non-oxidative metabolism of glucose. Thus, in a report published by G.A.R. with Sekine et al. [41], we demonstrated that the extractable enzymatic activity of lactate dehydrogenase (LDH) in fluorescence-activated cell-sorted (FACS) primary beta cells from the rat was remarkably low, and some 100-fold less than in the liver. This result was unexpected given the generally accepted view that LDH (encoded by the LDHA-D genes in rodents and humans) was a ‘housekeeping' enzyme in mammalian cells, responsible for ensuring rapid equilibration of pyruvate and lactate and required for lactate efflux in times of O₂ deprivation (as observed, for example, in exercising muscle). Whilst the above findings in primary beta cells were also true for the most well-differentiated rodent beta cell lines at the time, INS1 cells [42], less glucose-responsive lines (RINm5f, HIT-T15) showed dramatically higher LDH activities (Figure 2a). These findings suggested that, in the beta cell, pyruvate derived from glucose has one principal metabolic fate, i.e. decarboxylation by the pyruvate dehydrogenase (PDH) complex and oxidative degradation via the citrate cycle. Other measurements performed in the same study [41], and independently by Schuit et al. [43], demonstrated that >80% of glucose carbon is converted to CO₂ and H₂O in isolated rat beta cells, a remarkable feature, matched in the animal kingdom only by insect flight muscle [44]. Strikingly, inhibition of the respiratory chain in beta cells also inhibits glucose utilization, indicating that glycolysis and oxidative metabolism are obligatorily linked in these cells. Thus, neither the Pasteur effect (enhanced glucose utilization in the absence of O₂) nor the Crabtree effect (whereby high levels of ATP suppress oxidative metabolism and favor fermentation) [45], are observed in beta cells. Instead, high glycerol phosphate dehydrogenase (GPD2) [41,46,47], and to a lesser extent malate–aspartate shuttle [48] activities, provide efficient mechanisms to allow electrons derived from cytosolic NADH to enter the respiratory chain (Figure 1).

In an apparent challenge to these findings, working with Randle and others, Ashcroft et al. [16] reported glucose concentration-dependent lactate output from mouse islets. However, lactate production was completely refractory to the glucokinase inhibitor mannoheptulose (which blocks glucose oxidation completely). Malaisse et al. [49] also reported significant lactate output from whole intact rat islets, as did Tamarit-Rodriguez et al. [50] from beta cell-enriched ob/ob mouse islets. Tellingly, however, lactate output in this case did not increase at glucose 20 versus 3 mM glucose, despite a 2.5-fold and ∼4 fold increase in glucose utilization and oxidation, respectively. Even more strikingly, lactate and pyruvate transport activity [41], and expression of the monocarboxylate transporter MCT1 (Slc16a1) [51], were later found to be essentially absent from purified rat beta cells (Figure 2b). Taken together, these findings argue for a non-endocrine cell source of lactate from rodent islets. Correspondingly, lactate production was not detected from INS1 beta cells [41], nor the clonal sub-line, INS1 832/13 [48]. Ldha and Slc16a1 became the founder members of the so-called ‘disallowed' gene group in beta cells [52,53].

Recent findings suggest that the above scenario might be subtly different in human islets. LDHA is essentially absent from human beta cells (but expressed in alpha cells) [54,55]. On the other hand, and as first noted by Lawlor et al. [56], LDHB mRNA is clearly detectable in human beta (but not alpha) cells (Figure 2c) [55]. Accordingly, lactate generation from glucose, assessed using radiotracers and highly sensitive gas chromatography–mass spectrometry approaches is more substantial in human than in rodent islets [55]. However, and as for rodent islets [49], the interpretation of these experiments is complicated, since beta cells typically comprise only ∼35% of the total volume of human islet preparations (Figure 2c) [57] and LDHA (and B) levels are higher in contaminating ductal, endothelial, mesenchymal and other non-islet cells (Figure 2c). Moreover, in a recent study from Bilbao et al. [58], embryonic stem cell-derived beta cells, but not primary human islets, produced significant lactate from glucose. Nevertheless, the presence of LDHB in human beta cells at levels sufficient to equilibrate lactate and pyruvate may provide a ‘buffer' against extensive changes in redox potential and might limit the accumulation of reactive oxygen species (ROS). Interestingly, the latter, as well as contributing to the acute stimulation of insulin release by glucose (at least in rodents) [59], have been suggested recently [60] as promoters of rodent beta cell proliferation. On the other hand, lactate has recently been proposed to regulate cell division in immortalized cell lines by inhibiting the SUMO protease SENP1, thus stabilizing the anaphase-promoting complex (APC/C) to stimulate mitosis [61]. Although these studies have yet to be extended to the islet, the balance between intracellular lactate and ROS levels may conceivably contribute to inter-species differences in the potential for beta cell expansion.

In addition to influencing the metabolic fate of glucose, extinction of MCT1 in beta cells is also important to prevent pyruvate (or lactate) -induced insulin secretion during exercise [62,63]. This feature is unmasked in individuals carrying activating mutations in the SLC16A1 gene who, as a result, suffer from exercise-induced hyperinsulinemia and hypoglycemia. In rodents, the expression of LDHA is insufficient to allow significant flux from intracellular lactate, via pyruvate, towards mitochondrial oxidation even when MCT1 is up-regulated. Thus, lactate-induced secretion in rat islets is only evident after adenovirus-mediated overexpression of both Mct1 and Ldha [64]. Correspondingly, transgenic overexpression of Slc16a1 selectively in the beta cell facilitated pyruvate, but not lactate-induced, insulin secretion from mouse islets [63].

Although flux through PDH likely represents the chief metabolic fate for pyruvate in beta cells, other fluxes, notably ‘anaplerotic' input into the citrate cycle via pyruvate carboxylation (catalyzed by pyruvate carboxylase, PC) are also significant, and may represent as much as 40% of carbon flux from glucose into mitochondria [43,65]. The cataplerotic output of intermediates may then generate additional signaling molecules [23] — such as citrate or glutamate [66] (but see also [67]) — or reducing equivalents such a NADPH [68] generated via malate–pyruvate or isocitrate shuttles (Figure 1). Lipid coupling factors, notably monoacylglycerol [69], may also regulate the exocytotic machinery. Some, or all, of these molecules, may then contribute to the ‘amplification' of the effects of K_ATP channel closure on Ca²⁺-dependent insulin secretion [70]. Exploration of this possibility is, nevertheless, complicated, and inhibitors of PC, such as phenylacetic acid, also lower the cytosolic ATP/ADP ratio [71], indicating the importance of anaplerosis for the triggering (K_ATP channel-dependent) insulin secretion. Of note, the overall fluxes of mitochondrially derived metabolites are likely to be minor, given that the net efflux of glucose-derived carbons from beta cells other than as CO₂ (and possibly lactate) is limited (see above). Readers are referred to Prentki et al. [72], Maechler and Wollheim [66] and Merrins et al. [23] for a comprehensive survey of candidate signaling molecules for glucose-stimulated insulin secretion.

Activated Ca²⁺ influx, and oscillatory changes in cytosolic Ca²⁺, are fundamental features of glucose-stimulated insulin secretion (see above) [73,74]. Using the photoprotein aequorin [75], targeted to the mitochondrial matrix by in-frame fusion with the leader sequence of CoxVIII [76], we [77] demonstrated that increases in cytosolic calcium led to ∼10-fold higher increases in mitochondrial free calcium in rat-insulinoma-derived (INS-1 beta cells). Similarly, exaggerated mitochondrial (versus cytosolic) Ca²⁺ changes were later observed at high glucose [78]. These findings indicated that mitochondria are likely to serve as a sink, rather than a source of Ca²⁺ when beta cells are challenged at high glucose. As such, they were in line with the proposals of Denton (a former student of Randle) and McCormack [79], as discussed by Hansford [80], that stimulation of intramitochondrial dehydrogenases (pyruvate dehydrogenase, [PDH], NAD⁺-isocitrate dehydrogenase, NAD-ICDH [IDH2] and 2-oxoglutarate dehydrogenase [OGDH]) in mammalian cells enhances NADH production by the citrate cycle, ultimately driving increased ATP synthesis by the respiratory chain under circumstances where cytosolic ATP demand is increased [81].

To further explore a role for Ca²⁺ regulation of the intramitochondrial dehydrogenases in primary mouse beta cells, we [82,83] used a genetically-encoded ATP sensor based on the amino acid transporter GlnK1 (from the deep sea bacteria Methanococcus jannaschii), fused to GFP, alongside a red cytosolic Ca²⁺ probe. Simultaneously measuring plasma membrane potential using the cell-attached perforated patch, this approach also revealed that, after glucose challenge, an initial increase in ATP/ADP ratio was followed by transient decreases following action potential-induced Ca²⁺ spikes, likely reflecting an increase in ATP demand (for ion pumping, etc.) which transiently exceeded mitochondrial ATP synthesis, even after dehydrogenase activation by Ca²⁺. Cytosolic ATP/ADP then recovered gradually, possibly reflecting time-dependent activation of mitochondrial respiratory chain activity [84,85] (Figure 3). We were also able to show that Ca²⁺ transport converts a binary signal (cytosolic Ca²⁺ oscillations of near-unvarying amplitude) into a time-averaged change in amplitude which can be ‘read' by the TCA cycle [83].

Consistent with this view, silencing of the mitochondrial calcium uniporter MCU [86] in beta cells inhibited intramitochondrial Ca²⁺ increases and cytosolic ATP/ADP rises [82,87]. Inactivation of the Mcu gene selectively in the beta cell in mice by deletion of floxed alleles correspondingly impaired glucose-stimulated insulin secretion from primary islets [88]. Surprisingly, however, these animals remained normoglycemic, hinting at alternative pathways for mitochondrial Ca²⁺ uptake in vivo, or robust compensatory mechanisms after Mcu deletion.

The control of mitochondrial O₂ consumption and ATP synthesis in the beta cell thus appears, at different stages following stimulation by glucose, to be achieved by a combination of (1) ‘push' (i.e. increased substrate supply provided by GPD2 and citrate cycle-derived NADH), enhanced by Ca²⁺ activation of the intramitochondrial dehydrogenases and (2) ‘pull' (from lowered ATP/ADP ratios). The interplay between these is discussed below.

In displaying headroom for increases in ATP/ADP ratio as glucose concentrations rise, the beta cell differs from most other cell types since, in the latter, this ratio is usually held at high levels. In these cases, ATP synthesis is increased to meet cellular requirements, matching the demands of a given workload (e.g. for contraction, motility, ion pumping, biosynthesis, etc.). In contrast, the beta cell represents a special case, as a professional fuel sensor [20]. At fasting glucose concentrations (<5 mM), where the glycolytic generation of pyruvate from glucose is limited by the kinetic properties of the low affinity hexokinase, glucokinase (Michaelis constant, K_M > 10 mM), cytosolic ADP levels are higher (and ATP/ADP ratio lower, ∼3) in beta cells than most cell types, rising to ∼15 at high glucose [89]. Thus, at low glucose, mitochondria are in ‘state 3.5' [90], rather than state 4, such that substrate supply, rather than ADP availability, is limiting for respiratory chain activity. Hence, elevated glucose concentrations can enhance O₂ consumption and ATP synthesis, in the face of concomitant increases in cytosolic ATP/ADP (Figure 3, phase 1) [89,91–93]. Supporting this view, in recent elegant experiments using a microfluidic device, Schulze et al. [94] demonstrated that a step increase in glucose promotes a coincident increase in NAD(P)H, O₂ consumption and insulin secretion.

As suggested above, the bioenergetic situation in the beta cell is likely to be complex during glucose-induced oscillations in cytosolic ATP/ADP and Ca²⁺. Under these conditions, altered ADP supply to mitochondria, accelerated ATP consumption and an intrinsic glycolytic oscillator [95], may all contribute to fluctuations in ATP/ADP and O₂ consumption. In an important study, Luciani et al. [96] noted that Ca²⁺ appeared to shape NAD(P)H oscillations in islets, with forced Ca²⁺ increases elevating NAD(P)H at low glucose (via dehydrogenase activation) whilst lowering this parameter at high glucose, as inferred from changes in mitochondrial membrane depolarization. G.A.R.’s laboratory identified similar relationship based on cytosolic ATP/ADP measurements with cytosolic luciferase. Here, KCl-induced Ca²⁺ influx increased cytosolic ATP/ADP at low glucose, but decreased ATP/ADP at high glucose [97]. Changes in the opposite direction were observed when extracellular EGTA was used to suppress Ca²⁺ influx.

Li et al. [98] reported similar findings, including antiparallel changes in cytosolic Ca²⁺ and ATP at 20 mM glucose, providing further evidence that increased Ca²⁺ may stimulate O₂ consumption at Ca²⁺ peaks (ATP/ADP troughs). In contrast, Tanaka et al. [99], observed glucose-induced cytosolic ATP increases which preceded Ca²⁺ increases, whilst cytosolic ATP subsequently failed to oscillate in the face of Ca²⁺ oscillations. Whether these differences versus other reports may reflect the fact that the probe used measured ATP, rather than ADP, or that recordings were made over the whole islet, is unclear. For further discussion on the above the reader is referred to the excellent review by Bertram et al. [100].

Whether incretins affect mitochondrial metabolism directly is still contested. In studies on beta cell lines, Tsuboi et al. [101] demonstrated that glucagon-like peptide-1, (GLP-1) stimulated intracellular Ca²⁺ mobilization via protein kinase A (PKA)-stimulated phosphorylation of inositol-1,4,5 trisphosphate receptors, subsequently stimulating mitochondrial ATP synthesis and K_ATP channel closure. In contrast, Peyot et al. [102] observed no effects on intracellular ATP levels nor on O₂ consumption of treatment with GLP-1 or the analog exendin-4.

Finally, it should be noted that metabolic oscillations — however controlled — may represent a transitioning between two states: catabolic (O₂ consumption and ATP generation) and anaplorosis (output of signaling intermediates) [23].

Following the identification of multiple isoforms of the decarboxylase (E1) subunit of OGDH [103], to which Ca²⁺ binds [104], we demonstrated differential expression of these isoforms in different tissues with ∼30% of OGDH E1 mRNA encoding a Ca²⁺ insensitive form in islets (and brain) versus <5% in skeletal muscle and heart [105]. An important feature of the regulation of the citrate cycle dehydrogenases is that Ca²⁺ (synergistically with ADP and NAD⁺) lowers the K_M for substrate [105,106]. Thus, Ca²⁺ activation may be important to regulate substrate and product concentrations, as well as fluxes. In any case, where the Ca²⁺-insensitive variants predominate, accelerated conversion of 2-oxoglurarate to succinyl-CoA will be largely resistant to intramitochondrial Ca²⁺ increases. Consequently, isocitrate (and citrate) will be conserved for cataplerotic efflux from mitochondria, and thus the generation of potential coupling factors in the cytosol [23] (Figure 1).

Changes in mitochondrial shape and connectedness are also likely to impact the function of these organelles, and ultimately that of the ‘host' beta cell. Mitochondria exist as both threads (from the Greek, mitos) and grain- (chondros) like structures. The existence of the latter form is required for efficient turnover of mitochondria (mitophagy) and maintenance of a healthy mitochondrial population.

While mitochondrial fission precedes efficient turnover of aged or unhealthy mitochondria by mitophagy [107], maintaining a well-connected mitochondrial network is essential for metabolite and substrate exchange and thus optimal metabolic activity [108]. The balance between the two morphological configurations (fission and fusion) is tightly controlled depending on cell type, nutrient demand and extracellular/extraorganelle stimuli [109,110]. Abnormalities in this intricate balance have been implicated as an underlying cause for reduced mitochondrial function in metabolic diseases such as diabetes and obesity [108]. Indeed, beta cells from human islet donors with T2D display abnormal mitochondrial morphology with dilated or swollen mitochondria with reduced cristae and accompanied by diminished glucose-stimulated ATP/ADP ratio [111]. Moreover, and potentially complementing the actions of Ca²⁺ on intramitochondrial oxidative metabolism discussed above, Ca²⁺ also controls mitochondrial dynamics, at least in non-beta cells [112].

Given their role in integrating fuel metabolism, and the spatial requirements of both Ca²⁺ signals [113,114] and of ATP production [91] one might suspect that mitochondrial shape would be important for normal beta cell fuel sensing. These requirements may include allowing the formation of localized subcellular domains e.g. of ATP generation at the plasma membrane and the efficient operation of a ‘phophoenolpyruvate (PEP) cycle' [115] (Figure 1), allowing transfer energetic equivalents from GTP generated in the mitochondrial matrix towards ATP microdomains close to K_ATP channels at the plasma membrane [116].

Mitochondrial dynamics are orchestrated by the co-ordinated actions of membrane-associated dynamin-related proteins (DRPs), capable of oligomerization at the mitochondrial membrane and subsequent membrane remodeling [117]. Mitochondrial fission is mediated by dynamin-related protein 1 (DRP1) which forms a helix around the mitochondria, leading to membrane constriction and eventual scission. Mitochondrial fusion is mediated by mitofusins 1 and 2 (MFN1 and 2) and optic atrophy 1 (OPA1) localized on the outer mitochondrial membrane (OMM) and inner mitochondrial membrane (IMM), respectively. By dimerizing with their counterparts on neighboring mitochondria, MFN1 and 2 facilitate fusion by bringing the OMM into close proximity. This is followed by the interactions of long and short forms of OPA1 on the IMM of the opposing mitochondria leading to their fusion. The co-ordinated activity of these mitochondrial DRPs is mediated by cyclic GTP binding and hydrolysis that facilitates their oligomerization and disassembly [118]. Given the contribution of these events in controlling mitochondrial shape and network, the net balance between fission and fusion controls multiple aspects of mitochondrial physiology and quality control [119–121]. Mitochondrial fission by DRP1 allows for the segregation of depolarized mitochondria for clearance by mitophagy [107]. On the other hand, mitochondrial fusion allows for the maintenance of a homogenous mitochondrial network by ensuring uniform content distribution and exchange across the mitochondrial population [122–126]. Of note, the intermixing of mtDNA molecules with the mtDNA replication machinery allows for the maintenance of mtDNA integrity, while preventing mtDNA damage and accumulation of mutations [123,125].

Whereas the creation of a more highly fused network (achieved by inactivation of Drp1) leads to deficiencies in insulin secretion [127], we [128,129] have recently shown that fragmentation of mitochondria selectively in beta cells induced by disruption of the mitofusin genes (Mfn1 and Mfn2) leads to changes in beta cell function and to severe hyperglycemia in mice. Loss of MFN1/2 either during development [128] or in adult mice [129] was associated with a less highly interconnected mitochondrial network and also significant changes in the structure of the cristae. While deletion of Mfn1- of Mfn2- alone did not lead to mitochondrial defects, depletion of both Mfn1 and 2 caused significant hyperglycemia and a drastic reduction in mtDNA. In complementary studies, we [128] also showed that loss of mitofusin function in beta cells, accomplished by deleting the Mfn alleles one by one, was more closely correlated to mtDNA content than to network fragmentation per se.

In our gene dosage studies in mice, animals bearing only a single functional allele of Mfn2 displayed mitochondrial fragmentation while retaining normal glycaemic control, glucose-stimulated insulin release, and islet mtDNA content (Figure 4). On the other hand, mice bearing only a single allele of Mfn1 displayed mitochondrial network instability, mtDNA depletion, and significant hyperglycemia. These two distinct results implied that depletion of mtDNA content, rather than impaired mitochondrial architecture, leads to beta cell dysfunction. Moreover, our results clarified a long-standing challenge in the field of mitochondrial biology, distinguishing the relative roles for mitochondrial architecture and mtDNA copy number control by MFN1/2, which had previously been difficult to separate from one another. Thus, our studies established that the primary physiological importance of MFN1/2 in cellular metabolism in vivo is via regulation of mitochondrial genome stability, rather than mitochondrial structure.

Beta cell mass was only modestly affected in Mfn1–Mfn2 double knockout (DKO) mice [128,104] (and not detectably after deletion in adults [129,105]), though a slight increase in apoptosis was evident in both models. Beta cell identity, assessed through the expression of Slc2a2 and other signature genes, was only marginally affected, and we did not see an increase in the canonical dedifferentiation marker Aldh1a3 [130], or other ‘disallowed' genes [53]. Altered expression of genes in the autophagic pathway was, though, apparent [128].

Much more striking were alterations to the bioenergetic function of mitochondria. These included a drastic lowering in glucose-induced mitochondrial membrane potential changes and ATP synthesis after PdxCreER-mediated deletion in adult mice. Both cytosolic and, to an even greater degree, mitochondrial Ca²⁺ changes, were severely compromised in MFN1/2 DKO islets [129].

By encoding 13 essential subunits of the OXPHOS machinery, mtDNA enables individual mitochondria to regulate their own function depending on nutrient availability [131]. Consistent with reduced mtDNA, mtRNA, and protein levels of mitochondrially expressed proteins were markedly reduced in the MFN1/2 DKO islet. These findings suggested that the impact of Mfn1 and Mfn2 deletion on mtDNA content was due to decreased expression of the mitochondrial transcription factor, TFAM, and deleterious effects of mitofusin deletion changes were rescued in part by Tfam overexpression. Transcriptomic analysis suggested a post-transcriptional regulation of Tfam levels by Mfn1 and 2, as Mfn1/2 deletion did not affect Tfam mRNA levels. Also evident from our transcriptomic profiling of MFN1/2 DKO islets was an increase in the expression of Lonp1, a mitochondrially localized protease involved in the quality control of mitochondrial proteins [132]. Aberrant activation of LonP1 function, however, has been reported to cause a constellation of mitochondrial defects that include degradation of TFAM and destabilization of mtDNA nucleoids [133–135]. While it remains unclear how LonP1 regulates mitochondrial function in beta cells, the above results implicate Mfn1 and Mfn2 preservation of mtDNA content, in part through maintenance of TFAM.

Our studies add new and previously uncharacterized roles for mitofusins in beta cells. Together with previous work demonstrating beta cell dysfunction and glucose intolerance in vivo following loss of DRP1 or OPA1, these results demonstrate that the mitochondrial network balance is vital to promote beta cell health and function. However, several key questions remain. Our results suggest that MFN1/2 are essential for maintaining mtDNA content to promote mitochondrial respiration and insulin secretion. Furthermore, we find that GLP1R agonists can bypass mitochondrial defects to restore insulin secretion in MFN1/2-deficient islets, suggesting beta cells can activate alternative metabolic pathways to maintain insulin secretion during mitochondrial dysfunction (see next section). Notably, mice bearing Opa1-deficiency selectively in beta cells develop glucose intolerance and ultimately impaired beta cell replication [136], an effect not seen in MFN1/2-deficient islets. Moreover, mice with Drp1-deficiency in beta cells develop glucose intolerance that is speculated to be secondary to abnormalities in the amplifying pathway of insulin secretion and thought to be independent of mitochondrial respiration [127]. Taken together, these studies suggest the potential for a spectrum of cellular effects of altering the mitochondrial dynamics machinery in beta cells (both within and outside mitochondria) that remain to be harmonized. Additional clarity into the shared and distinct functions of the mitochondrial dynamics network will hopefully provide opportunities to better understand the linkages between beta cell mitochondrial structural and functional defects in T2D.

An exciting aspect of the above studies [129] was that the impact of deleting mitofusins on oral glucose tolerance (OGT) was greater than that on intraperitoneal tolerance (IGT). Given that the latter are driven by the release of incretin (e.g. GLP-1 and glucose-dependent insulinotropic peptide, GIP, and others) [26], we explored the effects of an incretin analog after mitofusin deletion. Exendin-4 almost completely rescued the abnormal glucose tolerance in vivo, and rescued defective glucose-induced insulin secretion in vitro. These changes were associated with normalized Ca²⁺ dynamics, including beta-cell beta cell connectivity defects. Pharmacological analysis suggested that rescue was due to engagement by cAMP of the Exchange Protein regulated (EPAC2) pathway, since it was still observed when PKA was inhibited, and was mimicked using an EPAC activator (Figure 5).

In addition to mitochondrial heterogeneity within individual beta cells [137], we have suggested [20] that differences in mitochondrial function or dynamics may contribute to the observed heterogeneity and connectivity between beta cells. Several recent reviews [138,139] have described these differences and we refer readers to those for a comprehensive discussion of this topic. Importantly, at the functional level, populations of both ‘leaders' from which Ca²⁺ waves appear to emanate [8,140–142], and highly connected ‘hub' [7,9] cells, each of which comprises ∼5% of all beta cells, play important and complementary roles in controlling islet wide Ca²⁺ dynamics and oscillatory insulin secretion. Johnston et al. [7] showed that mitochondrial hyper-polarisation in response to glucose was more marked in hubs versus the beta cell population as a whole. However, no differences between the structure of mitochondria (sphericity and mean length) were detected between the two populations. Differences in mitochondrial structure and function in ‘leader' versus other beta cells have not yet been explored. Likewise, Lu et al. [143] described an epigenomically discrete subset of beta cells in which mitochondrial genes were over-represented. Finally, Chabosseau et al. [140] revealed that several genes involved in mitochondrial function are differentially expressed between leader and follower beta cells.

In the healthy islet, glucose enhances mitochondrial oxidative metabolism, which is critical for the normal stimulation of insulin secretion. Incretins, including GLP1, potentiate these effects by only partly understood mechanisms. When mitochondria are fragmented as mitofusin levels are lowered, and conceivably in diabetes, they become dysfunctional as a result of being unable to maintain an adequate mtDNA content and hence respiratory chain function. GLP1R agonism rescues or bypasses these effects, ensuring normalized K_ATP channel-dependent calcium dynamics, beta-cell beta cell connectivity and insulin secretion. The impact of these changes on amplifying pathways is as yet undefined.

Whether differences may exist between the effects of GLP1R agonists which are biased for signaling via either cAMP-dependent pathways (including EPAC2) and through beta-arrestins [144] is also unexplored, but might represent a way of favoring the effects of these drugs to rescue deficiencies resulting from altered mitochondrial structure and function in T2D. Direct agonism of MFN2 [145] may provide a complementary therapeutic strategy.

Finally, the importance of mitochondrial function in the control of the secretion of other islet hormones (glucagon, somatostatin, etc.) represents an emerging area of interest, especially since changes in the expression of genes associated with mitochondrial metabolism in alpha cells are observed in both type 1 and type 2 diabetes [146].

G.A.R. has received grant funding from, and is a consultant for, Sun Pharmaceuticals Industries. S.A.S. has received grant funding from Ono Pharmaceutical Co., Ltd. and is a consultant for Novo Nordisk.

G.A.R. was supported by a Wellcome Trust Investigator Award (212625/Z/18/Z), UK MRC Program grant (MR/R022259/1), Diabetes UK Project grant (BDA16/0005485), CRCHUM start-up funds, an Innovation Canada John R. Evans Leader Award (CFI 42649), JDRF (JDRF 4-SRA-2023-1182-S-N), and CIHR (CIHR-IRSC:0682002550) project grants. S.A.S., B.A.K. and G.A.R. also acknowledge support from the NIH (R01 DK135268). S.A.S. acknowledges support from the JDRF (COE-2019-861), the NIH (R01 DK108921, R01 DK135032, R01 DK136547, U01 DK127747), and the Department of Veterans Affairs (I01 BX004444).

Guy A. Rutter: Conceptualization, Funding acquisition, Methodology, Writing — original draft. Vaibhav Sidarala: Writing — original draft, Writing — review and editing. Brett A. Kaufman: Writing — review and editing. Scott A. Soleimanpour: Writing — original draft, Writing — review and editing.

We thank Drs Stephen Ashcroft, Dick Denton, Mark Prentki and David Hodson for helpful discussion and Eleni Georgiadou for assistance with Figure 5.

AMPK

AMP-activated protein kinase

DKO

double knockout

DRPs

dynamin-related proteins

GLP-1

glucagon-like peptide-1

GPD2

glycerol phosphate dehydrogenase

IMM

inner mitochondrial membrane

LDH

lactate dehydrogenase

mtDNA

mitochondrial DNA

OGDH

2-oxoglutarate dehydrogenase

OMM

outer mitochondrial membrane

OPA1

optic atrophy 1

PC

pyruvate carboxylase

PDH

pyruvate dehydrogenase

PKA

protein kinase A

ROS

reactive oxygen species