Loss of Prohibitin Induces Mitochondrial Damages Altering β-Cell Function and Survival and Is Responsible for Gradual Diabetes Development

Phb2^fl/fl (18) and RipCre mice (19) were crossed to generate Phb2^fl/fl, RipCre⁺ mice designated as β-Phb2^−/−. In order to maintain single copy of the RipCre allele in homozygous β-Phb2^−/− knockout mice throughout the study, we systematically crossed heterozygous β-Phb2^-/wt (Phb2^fl/wt, RipCre⁺) males with Phb2^fl/fl (control) females. Because Phb2^fl/fl, RipCre⁺ mice (homozygous β-Phb2^−/− knockout) became diabetic at the age of 6 weeks, this strategy also avoided mating diabetic animals. As control mice, we used Phb2^fl/fl littermates in order to optimize standardization of the genetic background between the groups. Cre-mediated excision of Phb2 was assessed by PCR on genomic DNA extracted from isolated pancreatic islets using the primers 5′-ATCGTATTGGTGGCGTGCAGCA-3′ and 5′-AGGGAGGCTTGGTTTGAGGGGA-3′. Mice were maintained on a 12-h dark/light cycle and were allowed free access to standard laboratory chow (RM3-E-SQC #811181; SDS Diets, Essex, U.K.) and water. Mice were maintained in our animal facility according to procedures approved by the animal care and experimentation authorities of the Canton of Geneva.

Glucose (2 g/kg body weight) was administered intraperitoneally in 6 h–fasted mice before measurements of glucose levels on blood collected from tail vein at indicated times using a glucometer (Accu-Check; Roche Diagnostics, Rotkreuz, Switzerland). Hyperglycemia and diabetes were defined as blood glucose >11.1 mmol according to the criteria published by the American Diabetes Association (20). Plasma insulin levels from blood sampled by retro-orbital puncturing at time 0 and 15 min after glucose administration were determined using an ultrasensitive mouse insulin ELISA (Mercodia AB, Uppsala, Sweden). For plasma glucagon, blood was collected after 2-h fasting and 1-h refeeding as well as after 6-h fasting and 30 min after i.p. glucose (2 g/kg body weight) injection and glucagon levels determined by radioimmunoassay (RIA; GL-32K; Millipore, Billerica, MA).

Where indicated, mice were treated either with long-acting insulin (Levemir; Novo Nordisk, Gentofte, Denmark) injected subcutaneously twice per day (0.15 and 0.20 U in the morning and evening, respectively) or with leptin by using subcutaneous implantation of a 14-day osmotic pump (Alzet Model 1002; Alzet, Cupertino, CA) releasing 10 µg/day human leptin (Bachem, Bubendorf, Switzerland).

Pancreata were excised, weighed, fixed for 2 h in 4% paraformaldehyde, and finally embedded in paraffin. Sections of 5 µm separated by at least 250 µm were stained for insulin and glucagon using guinea pig anti-insulin (1:400) and mouse antiglucagon (1:500) primary antibodies as described (21). Fluorochrome-linked secondary antibodies were used for visualization, and images were captured by confocal microscopy (LSM 510 Meta; Carl Zeiss, Feldbach, Switzerland).

For assessment of α- and β-cell mass, sections at an interval of 250 µm throughout the pancreas were stained for glucagon and insulin, respectively, with the aforementioned primary antibodies. Horseradish peroxidase–conjugated secondary antibodies were used in order to reveal α- and β-cells by diaminobenzidine staining, and hematoxylin was used for counterstaining (21). Sections were scanned by digital microscopy (Nikon Coolscope; Nikon, Egg, Switzerland), quantification was achieved using ImageJ software (National Institutes of Health; http://rsb.info.nih.gov/nih-image/), and α- and β-cell mass was calculated as previously described (22).

For mitochondrial morphology, dispersed islet cells were allowed to adhere in culture to poly-L-lysine–coated slides before fixation and immunostaining with primary antibodies anti-TOM20 (1:100; Santa Cruz Biotechnology, Santa Cruz, CA) and anti-insulin (1:400; Sigma-Aldrich, St. Louis, MO), followed by Alexa 488– and -647–conjugated secondary antibodies, respectively. For each channel, z-axis stacks separated by 0.36 µm were acquired using a Zeiss LSM 510 Meta microscope with a 63 × 1.4 NA Plan Apochromat objective (Zeiss). Three-dimensional reconstruction as well as quantification of mitochondrial length was performed with the help of ImageJ software (National Institutes of Health) by manually measuring length of mitochondria in a randomly selected two-dimensional z-stack.

For ultrastructural analysis, isolated islets and whole pancreas of Phb2^fl/fl and β-Phb2^−/− mice were fixed in 2.5% glutaraldehyde in 0.1 mol/L phosphate buffer, by either immersion (islets, pancreas) or in situ perfusion (pancreas). After postfixation in 1% osmium tetroxide, all samples were processed for transmission electron microscopy as per standard procedures. Sections were analyzed in a CM10 Philips electron microscope (Philips).

Pancreatic islets were isolated by collagenase digestion as described (23) and cultured overnight in RPMI 1640 medium supplemented with 5% heat-inactivated FCS, 10 mmol HEPES, 1 mmol sodium pyruvate, 50 µmol β-mercaptoethanol, and antibiotics in nonadherent petri dishes. For static incubation studies, overnight cultured islets were preincubated with Krebs-Ringer bicarbonate HEPES (KRBH) buffer supplemented with 2.8 mmol glucose and 0.1% BSA for 1 h (23). Then, batches of 10 islets were hand-picked and incubated with basal 2.8 mmol and stimulatory 22.8 mmol glucose at 37°C for 1 h. At the end of the incubation period, supernatants were collected, and islets were resuspended in acid/ethanol for determination of insulin concentrations. For in vitro recombination experiments, islets isolated from adult Phb2^fl/fl mice were treated overnight either with control (Ad-LacZ)– or Cre-recombinase (Ad-RipCre)–expressing adenoviruses as described previously (24). The genomic deletion of Phb2 was confirmed 24 h after transduction by PCR as described previously (18). GSIS was tested, as mentioned earlier, 72 h after transduction. Additionally, in situ pancreatic perfusions were performed in anesthetized mice following a previously published procedure (23). Insulin concentrations in supernatants, perfusates, extracted islets, and pancreata were measured using RIA (Millipore). Total pancreatic glucagon contents were determined by glucagon RIA (Millipore).

Following overnight culture, islets were hand-picked and preincubated in KRBH at 2.8 mmol glucose for 1 h and then incubated at basal 2.8 mmol and stimulatory 22.8 mmol glucose for 15 min at 37°C before measurements of ATP levels as described previously (23).

For mitochondrial membrane potential, isolated islets were placed in 96-well plates for recording of rhodamine-123 signal (Fluostar Optima; BMG Labtech, Offenburg, Germany), maintained in KRBH with 2.8 mmol/L glucose before addition of stimulatory glucose concentration (final 22.8 mmol) according to published protocol (25).

For cellular Ca²⁺ measurements, isolated islets were allowed to adhere on poly-l-lysine–coated coverslips during overnight culture. Then, islets were incubated with 5 μmol/L Fura-2/AM in KRBH containing 2.8 mmol/L glucose and 0.02% pluronic acid for 30 min at 37°C, washed, and further incubated for 30 min in 2.8 mmol/L glucose KRBH. Coverslips were placed in a thermostatic chamber (Harvard Apparatus, Holliston, MA) maintained at 37°C, and images were acquired using an oil-immersion objective (40×) on an Axiovert microscope (Zeiss) as detailed previously (26).

Overnight cultured islets were washed once with PBS and lysed in standard RIPA buffer supplemented with protease inhibitor cocktail (Roche). Proteins were separated by SDS-PAGE and transferred onto nitrocellulose membranes. Primary antibodies used were directed against Phb1 and Phb2 (1:1,000; BioLegend, San Diego, CA), cleaved caspase-3 (1:1,000; Cell Signaling Technology, Danvers, MA), Opa1 (1:1,000; BD Biosciences, San Jose, CA), mitochondrial OXPHOS components (1:500; Mitosciences, Eugene, OR), and 4-hydroxynonenal (1:1,000; Abcam, Cambridge, U.K.). Isotype-matched HRP-conjugated secondary antibodies were used, and proteins were visualized by enhanced chemiluminescence (GE Healthcare, Glattbrugg, Switzerland). Densitometry analysis of different blots for mitochondrial complexes was performed with the help of Quantity One software (Bio-Rad, Hercules, CA).

TUNEL assay was performed on 15-min proteinase K–pretreated deparaffinized pancreas sections according to the manufacturer’s protocol by using the Apoptag Fluorescein apoptosis detection kit (S7110; Chemicon, Millipore, MA). Anti-insulin primary antibody (1:400) and its appropriate fluorochrome-coupled secondary antibody (1:500) were used to label β-cells. Nuclei were counterstained with DAPI.

Quantification of mtDNA copy number was achieved by quantitative PCR. Briefly, DNA was extracted from overnight cultured islets using the DNeasy Blood and Tissue kit (Qiagen, Düsseldorf, Germany) according to the manufacturer’s protocol. Nuclear and mtDNA copy numbers were assessed by quantitative PCR using primers targeted toward the COX1 gene (for mtDNA) and nuclear RNAse P (for nuclear DNA).

First, we documented the expression of prohibitins in human islets by immunoblotting (Fig. 1A). Next, we generated mice lacking β-cell Phb2 (β-Phb2^−/−) by crossing Phb2^fl/fl animals carrying a floxed Phb2 allele (18) with Tg(Ins2-cre)^1Herr mice, expressing Cre-recombinase under the control of an insulin promoter (19). PCR amplification across the Phb2 locus generated a Phb2-deleted fragment in islets of β-Phb2^−/− mice, but not in those of Phb2^fl/fl littermate controls, showing efficient Cre-mediated recombination (Fig. 1B). Consistent with genomic deletion, there was a near-total ablation of Phb2 protein in β-Phb2^−/− islets (Fig. 1C). The residual Phb2 expression observed in knockout islets is presumably accounted for by expression of the protein in non–β-cells and/or the persistence of the few β-cells that had escaped recombination. In islets isolated from β-Phb2^−/− mice, loss of Phb2 was accompanied by reduced levels of its homolog Phb1 (Fig. 1C), as previously observed (18). This reveals the interdependence of these proteins and the existence of functional prohibitin complexes in β-cells. Misexpression of Rip-Cre transgene in brain and hypothalamus has been noticed in a Tg(Ins2-cre)^25Mgn transgenic line (27), although not particularly in the Tg(Ins2-cre)^1Herr (19) used in the current study (28). Consistent with these reports, Phb2 loss in β-Phb2^−/− mice was restricted to pancreatic β-cells without apparent ablation in the brain, including the hypothalamus (Fig. 1D).

Monitoring of nonfasting glycemia in β-Phb2^−/− and littermate control Phb2^fl/fl mice showed normoglycemia in 4-week-old β-Phb2^−/− males. However, at the age of 6 weeks, knockout mice became hyperglycemic (>11.1 mmol), thereafter progressing to strong hyperglycemia (>25 mmol) and severe diabetes (weight loss) over a 3-week period (Fig. 1E). Similar phenotypes were observed in female β-Phb2^−/− mice (Supplementary Fig. 1). Recordings of body weight showed growth impairment starting at the age of 7 weeks, with no further weight gain compared with control littermates (Fig. 1F). Nonfasting hypoinsulinemia (54% less plasma insulin than control) displayed by 4-week-old β-Phb2^−/− mice (Fig. 1G) was further lowered at 10 weeks of age (88% less than age-matched control). Heterozygous Phb2 knockout (β-Phb2^wt/−) mice grew similarly to Phb2^fl/fl (data not shown) and had a normal life span, whereas homozygous β-Phb2^−/− mice died at the age of 12–15 weeks suffering of overt diabetes.

At 3 weeks of age, an intraperitoneal glucose tolerance test resulted in similar glycemic excursions in β-Phb2^−/− mice and control littermates (Fig. 2A). One week later, mice exhibited impaired blood glucose clearance along with normal fasting glycemia (Fig. 2B), indicating a prediabetic state. At 6 weeks of age, β-Phb2^−/− mice became severely diabetic with fasting hyperglycemia at ∼15 mmol/L (Fig. 2C). This timing of sequential alterations (normoglycemia followed by glucose intolerance and ultimately diabetes) was consistently observed in various litters. Thus, in this mouse model, the etiology of diabetes could be followed within a short time span of ∼3 weeks. Heterozygous β-Phb2^wt/− mice had normal glucose excursions even at the age of 10 weeks (Supplementary Fig. 2), indicating that the observed phenotype in β-Phb2^−/− mice was not mediated by the Rip-Cre allele present as a single allele in both heterozygous β-Phb2^wt/− and homozygous β-Phb2^−/− mice.

We further ascertained the possible impairment of insulin delivery suggested by low nonfasting plasma insulin levels (Fig. 1G). In 4-week-old β-Phb2^−/− mice, the plasma insulin levels observed 15 min after a glucose challenge were markedly lower (−50%) than in control mice (Fig. 2D). At the age of 6 weeks, β-Phb2^−/− mice exhibited very low fasting as well as glucose-induced plasma insulin levels (Fig. 2E). Therefore, diabetes development is first revealed in 4-week-old β-Phb2^−/− by glucose intolerance and lack of glucose-induced elevation of plasma insulin. Interestingly, 10-week-old β-Phb2^−/− mice exhibited hyperglucagonemia following a moderate 2-h fasting. Glucagon levels were further elevated in knockout animals upon refeeding (Fig. 2F), although not upon a glucose i.p. challenge (Supplementary Fig. 3).

At the age of 6 weeks, when Phb2 knockout animals became diabetic, we performed in situ pancreatic perfusions to test the β-cell response in islets maintained in their native pancreatic environment. This revealed a dramatic reduction of GSIS when pancreas of β-Phb2^−/− were compared with those of Phb2^fl/fl controls (Fig. 3A). In the former mice, first and second phases of insulin secretion were reduced by 76 and 78%, respectively, when compared with the corresponding phases of Phb2^fl/fl controls (Fig. 3B).

Because the chronic hyperglycemia observed by the age of 5 to 6 weeks might induce secondary changes perturbing islet function, we studied islets isolated from normoglycemic mice at the age of 3 to 4 weeks (Figs. 1E, 2A and B). Despite normoglycemia, insulin content in β-Phb2^−/− islets was decreased by 46% when compared with that of control islets (Fig. 4A). Islet function tested in static incubation revealed a markedly reduced GSIS in islets lacking Phb2 (Fig. 4B and C). When normalized per islet, the secretory response was 63% lower in β-Phb2^−/− islets compared with controls. Normalized per insulin content, the difference was −50%, indicating that, at this stage, the lower insulin content of β-Phb2^−/− islets did not account for their blunted glucose response. Accordingly, there was no significant difference between control and knockout islets when insulin secretion was stimulated by nonmetabolic KCl-induced cell depolarization (data not shown). This indicates that the exocytotic machinery downstream of Ca²⁺ signaling was preserved in β-Phb2^−/− islets. However, the Ca²⁺ rise secondary to glucose stimulation requires mitochondrial activation in terms of ATP generation. Hence, we measured intracellular Ca²⁺ concentration ([Ca²⁺]_i) in response to glucose stimulation in isolated islets. In comparison with littermate controls, β-Phb2^−/− islets exhibited a lower [Ca²⁺]_i rise upon glucose stimulation (Fig. 4D), while the response to KCl, used as a Ca²⁺-raising agent, was similar in the two groups (Supplementary Fig. 4).

Additionally, we acutely depleted Phb2 in vitro in β-cells by treating nonrecombined Phb2^fl/fl islets isolated from adult mice (age 10–14 weeks) with adenovirus expressing Cre recombinase (Fig. 5A and B). Compared with control islets transduced with Ad-LacZ virus, we observed impaired GSIS in recombined Ad-RipCre treated islets 72 h after viral transduction (Fig. 5C). These data demonstrate that Phb2 is required for proper metabolism-secretion coupling in β-cells.

We next studied mitochondrial morphology in animals aged 3 to 4 weeks. Three-dimensional reconstructions of confocal microscopy z-stacks images of mitochondrial reticulum through entire β-cells are shown in Fig. 6A and Supplementary Fig. 5A and B. Normal mouse β-cell displayed tubular interconnected mitochondria. On the contrary, absence of Phb2 in β-cells resulted in fragmented mitochondrial network, revealing an average mitochondrial length ∼50% shorter compared with control β-cells (Fig. 6B). In accordance with this fragmented mitochondrial network, we observed excessive proteolytic cleavage of the long isoforms (L1 and L2) of the mitochondrial inner membrane fusion protein Opa1, resulting in the accumulation of the short isoform S5 (Fig. 6C). In spite of these changes, electron microscopy revealed a comparable ultrastructural appearance of mitochondria in the β-cells of 4-week-old control Phb2^fl/fl and knockout β-Phb2^−/− mice, whether these cells were studied in isolated islets (Fig. 6D and E) or intact pancreas (not shown). Similar observations were made in β-cells of newborn, 2-week-, and 6-week-old mice (not shown). These pictures (Fig. 6A, D, and E) show that mitochondrial fragmentation is not necessarily associated with alteration of mitochondrial ultrastructure.

In neuronal mitochondria, Phb2 is necessary for the maintenance of mtDNA, probably requiring Opa1-dependent mitochondrial fusion (29). In this context, we quantified mtDNA copy number normalized to nuclear DNA. This revealed a 47% reduction of the mitochondrial genome in Phb2 null islets versus controls (Fig. 6F), which was accompanied by lower levels of mtDNA-encoded subunit of complex IV of the respiratory chain at the protein level (Fig. 6G and H). Alterations in the electron transport chain might favor production of reactive oxygen species (30,31). However, we recorded similar levels of 4-hydroxynonenal, an aldehydic product of lipid peroxidation commonly enhanced during oxidative stress (Supplementary Fig. 6A and B).

The mitochondrial function was assessed by measuring mitochondrial membrane potential and ATP. Upon glucose stimulation, mitochondria of Phb2 null β-cells exhibited weak hyperpolarization compared with their control counterparts (Fig. 6I). This was translated into impaired ATP generation after glucose stimulation. At basal 2.8 mmol/L glucose, Phb2^fl/fl and β-Phb2^−/− islets had similar concentrations of ATP (29.4 ± 9.2 and 28.2 ± 12.2 pmol/µg islet proteins, respectively). Upon 22.8 mmol glucose stimulation, we measured a 30% rise in ATP for control islets versus only 7% for β-Phb2^−/− islets (Fig. 6J), which might be sufficient for the blunted elevation of [Ca²⁺]_i (Fig. 4D). These data show that loss of Phb2 induced alterations in mitochondrial morphology and function, resulting in an impaired metabolic response.

The defective GSIS observed in β-Phb2^−/− mice might be caused by physical or functional loss of β-cells or both. Hematoxylin and eosin staining of pancreas of 4-week-old knockout mice revealed absence of infiltration of immune cells within and around islets (Supplementary Fig. 7). At the same age, immunohistochemistry showed that most islets of β-Phb2^−/− mice exhibited reduced size with preserved architecture (i.e., glucagon-positive α-cells at the periphery and insulin-producing β-cells forming the core of the islet) (Fig. 7A). At the age of 10 weeks, however, islets were completely disorganized in β-Phb2^−/− mice, with only few β-cells intermingled with α-cells (Fig. 7B). Quantification of islet β-cell mass in β-Phb2^−/− mice revealed that β-cell loss amounted to 35% at the age of 4 weeks and >90% at the age of 10 weeks compared with age-matched controls, consistent with the corresponding alterations in pancreatic insulin content (Fig. 7C and D). In 10-week-old β-Phb2^−/− mice, we observed an expanded α-cell mass compared with that seen in controls (Fig. 7E). This was substantiated by higher pancreatic glucagon content in β-Phb2^−/− mice at 10 weeks of age compared with control animals (Fig. 7F).

Deficient β-cell function at 4 weeks of age, followed by loss of β-cell mass at 10 weeks in β-Phb2^−/− mice, might explain the development of diabetes. In order to test if the lack of insulin delivery was a key contributor to the disease, β-Phb2^−/− animals were treated with insulin. The daily administration of insulin to β-Phb2^−/− diabetic mice, which was initiated at 8 weeks of age, fully corrected glycemia (Fig. 7G) and provided for maintenance of control body weights (Fig. 7H) for the 10-week period of the treatment. Untreated β-Phb2^−/− mice were killed at 11 weeks of age because of a severe diabetic state with dramatic loss of body weight. Incidentally, the rescue by insulin treatment of multiple alterations of the β-Phb2^−/− mice shows the absence of a hypothalamic contribution to the observed phenotype. Leptin therapy in diabetic mice has been shown to normalize glycemia through suppression of hyperglucagonemia (32). In view of the higher α-cell mass and pancreatic glucagon of β-Phb2^−/− mice (Fig. 7E and F), we tested leptin treatment in these animals. Fig. 7I shows that subcutaneous delivery of leptin over a 10-day period partially corrected glycemia in 8-week-old diabetic β-Phb2^−/− mice.

Islets isolated from 4-week-old β-Phb2^−/− mice exhibited a marked upregulation of caspase-3 cleavage compared with age-matched controls (Fig. 8A). Additionally, these cells were more susceptible to apoptotic stimuli, such as staurosporine (Fig. 8A). On pancreas sections, we typically observed one to three apoptotic β-cells by TUNEL assay in β-Phb2^−/− islets, while we detected none in control islets (Fig. 8B). Given that β-Phb2^−/− mice maintained a sufficient β-cell mass up to the age of 4 weeks, before the rapid development of diabetes, data suggest the involvement of some compensatory mechanism.

Preservation of ∼65% of the β-cell mass in 4-week-old β-Phb2^−/− mice (Fig. 7C) despite active apoptosis (Fig. 8A and B), prompted us to investigate β-cell proliferation. Quantification of Ki-67–positive β-cells in control mice showed the expected 1% proliferating β-cells at the age of 4 weeks (Fig. 8C). Surprisingly, we observed an increased number of Ki-67–positive nuclei within insulin-positive cells of β-Phb2^−/− mice (Fig. 8C). This resulted in a 2.5-fold β-cell proliferation increment in β-Phb2^−/− mice versus control (Fig. 8D). However, at the age of 10 weeks, the 90% reduction in β-cell mass of β-Phb2^−/− mice (Fig. 7C) suggested loss of such a compensatory proliferation.

The current study documents the expression of prohibitins and their importance for β-cell function and survival. Ablation of Phb2 in mouse β-cells sequentially resulted in impaired mitochondrial function and insulin secretion, loss of β-cells, progressive alteration of glucose homeostasis, and ultimately severe diabetes. Defective insulin supply was contributed by both β-cell dysfunction and apoptosis, suggesting a pivotal role for Phb2 in maintenance of the β-cell integrity. At the molecular level, we observed that deletion of Phb2 caused mitochondrial abnormalities such as reduction of mtDNA copy number and complex IV levels. Our β-Phb2^−/− mice share some phenotypic similarities with β-cell–specific frataxin knockout. Frataxin is located in the mitochondrial matrix, controlling iron-sulfur cluster assembly (33). Mice lacking frataxin in β-cells are born healthy but subsequently develop glucose intolerance and then diabetes by the age of 9 months, explained by oxidative stress, apoptosis, and then reduced islet mass (31). Phb2 deficiency caused more rapid development of diabetes across the age of 3 to 6 weeks, also accompanied by β-cell loss but without apparent oxidative stress.

Phb2-null β-cells exhibited short, fragmented, and globular mitochondria, which, however, retained a normal ultrastructural appearance, as judged by the persistence of a double membrane boundary and numerous thin elongated cristae. This was nevertheless associated with mitochondrial dysfunction, as shown by reduced glucose-induced ATP production. Phb2 knockout β-cells had accelerated proteolytic degradation of L-Opa1 isoforms. Indeed, we observed the accumulation of the short Opa1 isoform S5, when a balanced proportion of short and long forms is necessary for the proper function of this protein (16). Deletion of Opa1 in mouse β-cells decreases the activity of electron transport chain complex IV and alters ATP generation (9). The present data indicate that a selective loss of L-Opa1, and/or the accumulation of its short isoform, results in mitochondrial pattern distinct from complete ablation of Opa1 (9). Fragmented mitochondrial pattern has been documented in β-cells of human type 2 diabetic subjects and animal models for diabetes (8,34,35). However, it was unclear from these studies whether the alterations in mitochondrial morphology were the cause or consequence of diabetes. Strikingly, alterations of mitochondria appeared well before the onset of diabetes in β-Phb2^−/− mice. Therefore, our data favor a role for Phb2 in Opa1-dependent mitochondrial fusion for β-cell function and morphology.

We measured lower mtDNA copy number and complex IV levels in β-Phb2^−/− islets. In Opa1-null islets, mtDNA copies are unchanged (9). However, in neurons lacking Phb2, destabilization of L-Opa1 is associated with progressive loss of the mitochondrial genome (29), pointing to a complex equilibrium between the Opa1 isoforms for the maintenance of mtDNA integrity. In β-Phb2^−/− islets, degradation of L-Opa1 accompanied by a decline in mtDNA might explain the marked reduction in the glucose response, which was more severe than in the complete absence of Opa1 (9).

Mitochondrial abnormalities were present in the β-cells of β-Phb2^−/− mice, rendering these animals severely diabetic by the age of 6 weeks. The lower insulin release of β-Phb2^−/− mice was also contributed by reduced β-cell mass, although at the age of 4 weeks, the secretory response was decreased independently of insulin content and specifically for mitochondrion-dependent secretagogue. Glucose intolerance appeared at this age of 4 weeks when β-cell mass in β-Phb2^−/− mice was still ∼65% of control animals, whereas hyperglycemia developed 2 weeks later. Clinical data have documented the maintenance of glucose tolerance in healthy donors of functional islets who underwent partial pancreatectomy (36). Interestingly, a recent study in humans reported that diabetes appears after a reduction in β-cell mass of ∼65%, while postchallenge glucose excursions in prediabetic subjects exhibit glucose intolerance before this critical threshold is reached (37). Therefore, within a 2- to 3-week period, the β-Phb2^−/− mouse model recapitulates the progressive stages of human diabetes, with progression from β-cell impairment associated with glucose intolerance to diabetes with fasting hyperglycemia.

In mouse embryonic fibroblasts lacking Phb2 (18), as well as in Opa1-null β-cells (9), apoptosis is not increased (unless stimulated by extrinsic stimuli), although the proliferative capacity of β-cells is lost. In marked contrast, the 4-week-old β-Phb2^−/− mice exhibited a higher β-cell proliferation, which partially compensated the enhanced apoptosis during the first weeks of life. Apoptotic TUNEL-positive β-cells were regularly observed in islets of β-Phb2^−/− mice, while such events were extremely rare in control mice. This suggests that β-Phb2^−/− mice gradually lost β-cells by apoptosis over a period of ∼10 weeks. Enhanced apoptosis mediated by cleavage of caspase-3 was a likely consequence of Phb2 deletion rather than Opa1 degradation, since caspase-3 is not activated in Opa1-null β-cells (9).

In this study, we also observed the expansion of the α-cell mass and pancreatic glucagon content in 10-week-old β-Phb2^−/− mice. In diabetic patients, loss of β-cells accompanied by an increase in α-cell mass has been documented and might contribute to hyperglucagonemia, precipitating hyperglycemia because of increased hepatic glucose production (38). Interestingly, after 2 h of fasting, β-Phb2^−/− mice exhibited hyperglucagonemia, which was further enhanced following food ingestion. Moreover, leptin treatment partially corrected glycemia in diabetic β-Phb2^−/− mice, in accordance with a glucagon-suppressor effect of leptin (38). These findings support the growing interest in α-cell as a critical therapeutic target for the treatment of diabetes.

β-Phb2^−/− mice represent a unique model of spontaneous diabetes development, through a series of molecular events appearing over a 3-week period, not requiring the administration of toxic diets or chemicals. Overall, our data demonstrate that Phb2 is essential for the function and survival of β-cells. Phb2 regulates mitochondria by preserving some of their key components such as mtDNA, respiratory chain subunits, and the morphology regulator L-Opa1. Phb2 ablation impairs mitochondrial activation, rendering β-cells unresponsive to glucose and ultimately leading to apoptosis, which promotes β-cell loss and diabetes.

This work was supported by the Swiss National Science Foundation (grants 310030B_135704 to P.Ma. and 310030_141162 to P.Me.) and the State of Geneva.

No potential conflicts of interest relevant to this article were reported.

S.S. conducted experiments, analyzed data, and wrote the manuscript. F.T., L.S., and P.Me. developed specific techniques and analyzed corresponding data. C.M., P.L.H., and T.L. generated transgenic animals essential for this study. A.G. generated data. P.Ma. supervised the project, analyzed data, and wrote the manuscript. P.Ma. 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.

The authors thank Thierry Brun (University of Geneva) and Mauro Corrado (University of Geneva) for advice and helpful discussions, and Mrs. Dorothée Caille (University of Geneva) for processing the EM samples.