We have previously identified transcription factor B1 mitochondrial (TFB1M) as a type 2 diabetes (T2D) risk gene, using human and mouse genetics. To further understand the function of TFB1M and how it is associated with T2D, we created a β-cell-specific knockout of Tfb1m, which gradually developed diabetes. Prior to the onset of diabetes, β-Tfb1m−/− mice exhibited retarded glucose clearance owing to impaired insulin secretion. β-Tfb1m−/− islets released less insulin in response to fuels, contained less insulin and secretory granules and displayed reduced β-cell mass. Moreover, mitochondria in Tfb1m-deficient β-cells were more abundant with disrupted architecture. TFB1M is known to control mitochondrial protein translation by adenine dimethylation of 12S ribosomal RNA (rRNA). Here, we found that the levels of TFB1M and mitochondrial-encoded proteins, mitochondrial 12S rRNA methylation, ATP production and oxygen consumption were reduced in β-Tfb1m−/− islets. Furthermore, the levels of reactive oxygen species (ROS) in response to cellular stress were increased whereas induction of defense mechanisms was attenuated. We also show increased apoptosis and necrosis as well as infiltration of macrophages and CD4+ cells in the islets. Taken together, our findings demonstrate that Tfb1m-deficiency in β-cells caused mitochondrial dysfunction and subsequently diabetes owing to combined loss of β-cell function and mass. These observations reflect pathogenetic processes in human islets: using RNA sequencing, we found that the TFB1M risk variant exhibited a negative gene-dosage effect on islet TFB1M mRNA levels, as well as insulin secretion. Our findings highlight the role of mitochondrial dysfunction in impairments of β-cell function and mass, the hallmarks of T2D.
Failure to secrete sufficient amounts of insulin is ultimately the culprit in the development of type 2 diabetes (T2D), one of the most devastating diseases of our time (1). Fully functional β-cells in a sufficient number is a prerequisite of proper insulin secretion. There is ample evidence for perturbations of both function and mass of β-cells in T2D (2,3). These processes may be influenced by altered mitochondrial function, which has been shown to be compromised in islets from patients with T2D (4). In fact, we recently reported that a variant of TFB1M, which encodes a protein essential for mitochondrial function, is associated with future risk of T2D (5).
The β-cell has developed an elaborate mechanism, known as stimulus-secretion coupling, whereby a rise in extracellular glucose, occurring after a meal, is translated into intracellular signals that permit the proper amount of insulin to be released. Circulating insulin then binds to receptors in its main target tissues—skeletal muscle, adipocytes and liver—and promotes glucose uptake and anabolism. The transport of glucose into the β-cell activates a complex metabolic pathway that results in ATP production via electron transport and oxidative phosphorylation (OXPHOS). The formation of ATP leads to the closing of ATP-dependent K+ (KATP) channels (6). Thereby the plasma membrane is depolarized, and subsequently intracellular Ca2+ rises, triggering exocytosis of insulin (7,8).
Insulin secretion is thus triggered but also maintained via the generation of signals from the β-cell mitochondria (9,10). The molecular machinery in these organelles is controlled by a unique interplay of nuclear- and mitochondrial-encoded genes. Mitochondrial DNA encodes 13 essential proteins that are part of complexes I, III, IV, which carry out cell respiration, and ATP synthase. Transcription of mitochondrial genes is controlled by nuclear-encoded factors: transcription factor A mitochondrial (TFAM) and transcription factors B1 and B2 mitochondrial (TFB1M and TFB2M). TFAM has a dual function: it is involved in the packaging of mitochondrial DNA into nucleoids and activation of transcription (11). Although TFB1M [also known as S-adenosylmethionine-6-N′,N′-adenosyl(rRNA)-dimethyltransferase 1 mitochondrial (mDMAT1)] was originally thought to serve as a transcription factor along with TFB2M, it is now established that it mainly functions as a dimethyltransferase (12). In mitochondria, TFB1M dimethylates two adjacent adenine residues in the hairpin loop at the 3′ end of 12S rRNA; this modification is a mandatory step in the biogenesis of the small subunit of the mitochondrial ribosome (13,14). Deficient 12S rRNA methylation results in destabilized ribosomes in the mitochondria, leading to severely impaired translation of mitochondrial proteins (12).
The discovery that a common variant of TFB1M is associated with impaired insulin secretion and increased risk of future T2D recently highlighted the pathogenetic relevance of mitochondrial protein synthesis for the human disease (5). Carriers of the risk variant had reduced levels of TFB1M and mitochondrial-encoded respiratory proteins in islets as well as impaired insulin secretion in vivo and in vitro. Furthermore, we have also shown that TFB1M deficiency causes mitochondrial dysfunction and impaired insulin secretion in mice with a general heterozygous deficiency of Tfb1m (Tfb1m+/−), as well as in rat clonal β-cells (5). These findings show that mitochondrial dysfunction is a causal pathogenetic process in the common form of human T2D. The underlying mechanisms and the precise role of TFB1M in pancreatic β-cells, however, remain unclear.
To elucidate the mechanisms underlying mitochondrial dysfunction and the subsequent development of T2D owing to loss of TFB1M, we examined the role of Tfb1m in murine β-cells. We show that conditional targeting of Tfb1m in β-cells results in mitochondrial dysfunction, which leads to impaired stimulus-secretion coupling, perturbed insulin secretion, energy failure, increased production of ROS, progressive β-cell loss and ultimately development of diabetes. Importantly, we also show that human carriers of the TFB1M risk variant have reduced expression of TFB1M mRNA in islets as well as lower insulin secretion. These findings further support the relevance of our murine findings for the development of the human disease.
Expression and function of TFB1M are abrogated in islets from β-Tfb1m−/− mice
To understand the role of TFB1M in pancreatic β-cells, we inactivated the gene specifically in these cells by expressing the Cre recombinase under control of the rat insulin 2 promoter (Rip-Cre) in pancreatic β-cells in mice homozygous for a loxP-flanked exon 3 of the Tfb1m locus. We first measured the mRNA from islets isolated from β-Tfb1m−/− and control mice and found that Tfb1m mRNA levels from the exon boundary 2–3 and 3–4 were dramatically reduced in β-Tfb1m−/− islets compared with control islets (Fig. 1A); the expression of other exons of the transcript was moderately reduced in β-Tfb1m−/− islets. As some rat insulin promoters in transgenic mice have been found to confer Cre expression in the hypothalamus (15), we examined the expression of Tfb1m mRNA in the hypothalamus of β-Tfb1m−/− mice and found its levels to be similar with controls (Fig. 1B).
To further confirm these findings, we measured TFB1M protein levels and found these to be markedly lower, but not completely absent, in islets from pre-diabetic β-Tfb1m−/− mice (Fig. 1C). This finding indicates that Cre-mediated recombination has occurred in the majority of β-cells; the remaining expression can be attributed to either a minor population of β-cells, in which recombination has not occurred, or non-β-cells, which constitute 25–35% of islets cells. The abundance of TFB1M protein in the hypothalamus was similar in β-Tfb1m−/− and control mice (Fig. 1D).
Given that TFB1M mainly functions as a dimethyltransferase, we then investigated the impact of β-cell deficiency of Tfb1m on mitochondrial ribosomes by measuring methylation of 12S rRNA, using a primer extension assay (12). Indeed, we found a significant decrease in the methylation of 12S rRNA in islets from pre-diabetic β-Tfb1m−/− mice (Fig. 1E and F). In contrast, 12S rRNA adenine dimethylation in the hypothalamus of β-Tfb1m−/− mice was unchanged (data not shown). Thus, expression and function of TFB1M were abrogated in β-Tfb1m−/− islets.
Expression of both mRNA and proteins that are important for mitochondrial function is perturbed in β-Tfb1m−/− mice
Given that TFB1M has also been suggested to function as a transcription factor (16,17), we analyzed the expression of genes essential for mitochondrial function, using Q-PCR on total cDNA in islets from pre-diabetic β-Tfb1m−/− mice. Our results show that the mRNA expression for nuclear-encoded genes essential for mitochondrial function, e.g. OXPHOS, was unaltered (Fig. 2A). Furthermore, mRNA levels for mitochondrial-encoded genes (Fig. 2B) were largely unaffected, with the exception of 16S rRNA and Nd1, which were increased 1.7-fold (P < 0.05) and 1.3-fold (P < 0.05), respectively. In contrast, the expression of Cytb was reduced 1.4-fold (P < 0.05; Fig. 2B).
Given the critical role of TFB1M for mitochondrial protein translation, we then examined the abundance of mitochondrial proteins. We found that the abundance of the mitochondrial-encoded subunits of complex I and IV, ND1 and COX1, as well as a nuclear-encoded subunit of complex I, NDUFB8, was significantly reduced in islets from pre-diabetic β-Tfb1m−/− mice (Fig. 2C and D). In contrast, protein levels of the nuclear-encoded ATPA subunit (FOF1 ATP synthase, subunit α) and complex II were increased (Fig. 2C and D). These data show that the lack of TFB1M impaired protein translation in β-cell mitochondria before the development of diabetes.
Mitochondrial dysfunction in β-Tfb1m−/− mice
Next, we determined whether mitochondrial function was altered in β-Tfb1m−/− islets, using a dye reflecting mitochondrial membrane polarization (tetramethyl-rhodamine methylester, TMRM). We found that upon stimulation with 16.7 mm glucose, the maximal decrease in TMRM fluorescence intensity from baseline was reduced by 42% in β-Tfb1m−/− islets compared with controls (P < 0.01, Fig. 3A and B). There was, however, no difference in the increase of TMRM fluorescence intensity compared with baseline upon stimulation of control and β-Tfb1m−/− islets with the uncoupler carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP).
Following fuel-stimulated polarization of the inner mitochondrial membrane, as reflected by our TMRM measurements, O2 is consumed when H2O is reduced by complex IV in the respiratory chain. Therefore, to further examine mitochondrial function when TFB1M is deficient, we measured O2 consumption, which reflects the rate of OXPHOS in islets from pre-diabetic β-Tfb1m−/− mice. The control islets showed a progressive increase in oxygen consumption rate (OCR) after a rise in glucose from 2.8 to 16.7 mm (Fig. 3C). In contrast, β-Tfb1m−/− islets showed a lower OCR at the basal glucose level and failed to increase it to the same extent as control islets in response to an elevation of the glucose concentration (Fig. 3C). Consequently, the area under the curve (AUC) for OCR upon a rise in glucose was reduced in β-Tfb1m−/− islets compared with control islets (Fig. 3D).
The final step in mitochondrial metabolism is production of ATP, which is involved in both triggering and amplification of glucose-stimulated insulin secretion (9,10). We examined whether Tfb1m deficiency affected the kinetics of mitochondrial ATP production in response to the mitochondrial fuels glutamate and succinate before development of diabetes. In addition, islet ATP content was determined 30 min after incubation with glucose and inhibitors of mitochondrial ATP synthesis. We found a pronounced decrease in the rate of mitochondrial ATP production in β-Tfb1m−/− islets (Fig. 3E). Consequently, β-Tfb1m−/− islets showed reduced ATP content after incubation in 2.8 or 16.7 mm glucose (Fig. 3F). However, the reduction of ATP content in the presence of 16.7 mm glucose and a mixture of 40 µm of the uncoupler 2,4-dinitrophenol (2,4-DNP) and 4 µg/ml oligomycin, which blocks mitochondrial ATP production, was more pronounced in control versus β-Tfb1m−/− islets: 4.4-fold (P < 0.001) and 2.2-fold (P < 0.01), respectively, compared with 16.7 mm glucose alone. This result suggests that glycolytic production of ATP is increased in β-Tfb1m−/− islets, serving to compensate for mitochondrial dysfunction. Accordingly, the mRNA levels for lactate dehydrogenase A (Ldha), essential for glycolytic and anaerobic glucose metabolism, were increased 1.2-fold (P < 0.05) in β-Tfb1m−/− islets (Supplementary Material, Fig. S1). These results show that disruption of Tfb1m in β-cells leads to impaired mitochondrial ATP production owing to perturbed OXPHOS.
Mitochondrial ultrastructure is disrupted in Tfb1m-deficient β-cells
It has previously been observed that Tfb1m deficiency leads to increased mitochondrial number and altered mitochondrial structure in cardiac muscle cells (12). To examine this possibility in pancreatic islets, we analyzed mitochondrial ultrastructure by electron microscopy (see Supplementary Material, Fig. S2, for the procedure of calculating mitochondrial area). Mitochondria in β-cells from control islets were of the common crista type: they exhibited a typical elongated or rounded shape with crest-like protrusions, bright matrix and mutually parallel cristae (Fig. 4A and E). In contrast, mitochondria in β-cells from pre-diabetic β-Tfb1m−/− mice displayed an abnormal shape: the cristae were less distinct with mitochondrial clumping (Fig. 4B–D, F; Supplementary Material, Fig. S2). The cross-sectional area covered by mitochondria increased significantly from 6.8 µm2 in control β-cells to 19.3 µm2 in β-Tfb1m−/− β-cells (Fig. 4G). This increase was not accompanied by any change in total β-cell cross-sectional area; the ratio of mitochondrial cross-sectional area and the total cell cross-sectional area in β-cells was 2.7-fold higher in β-Tfb1m−/− compared with control (P = 4.5E-12). These findings demonstrate that altered action of TFB1M in β-cells had a profound effect on mitochondrial structure, resulting in increased mitochondria-covered area and perturbed architecture.
Whole-body metabolism and insulin secretion are perturbed in β-Tfb1m−/− mice
To further investigate our previous findings about a possible role of TFB1M in diabetes development (5), we then examined whole-body metabolism. We found that body weight did not differ between early-stage β-Tfb1m−/− (age: 76 ± 17 days) and control mice but late-stage β-Tfb1m−/− mice (age: 149 ± 19 days) had reduced body weight compared with controls (Fig. 5A). Plasma glucose levels were measured longitudinally in non-fasted control and β-Tfb1m−/− mice: at late stage, virtually all β-Tfb1m−/− mice had developed diabetes (Fig. 5B). Indeed, at that time point, the plasma insulin concentration in β-Tfb1m−/−mice was significantly lower than that of control mice in the non-fasted state (Fig. 5C; in 3 of 8 β-Tfb1m−/− mice, plasma insulin was below the detection limit of the assay). Accordingly, β-Tfb1m−/−mice showed clinical symptoms of diabetes at the late stage (age: 149 ± 19 days), i.e. ketosis, cachexia, polyuria and polydipsia.
To further elucidate the mechanism of hyperglycemia, we assessed glucose homeostasis in the β-Tfb1m−/− mice before diabetes onset (age: 76 ± 17 days) by an intraperitoneal glucose tolerance test (IPGTT): β-Tfb1m−/− mice showed retarded clearance of glucose (Fig. 5D) and blunted insulin secretion (Fig. 5E). Glucose tolerance was similar in Tfb1mloxP/loxP (β-Tfb1m+/+; control) and Tfb1mRip−Cre/− (Rip-Cre-transgenic) mice (Supplementary Material, Fig. S3). These findings indicate that β-Tfb1m−/− mice gradually lose the capacity to secrete insulin and progress from glucose intolerance to overt diabetes.
To understand the defective insulin secretion in vivo, we measured insulin secretion from isolated islets prior to diabetes development in β-Tfb1m−/− mice. We postulated that Tfb1m deficiency in β-cells would exert multiple effects on insulin secretion owing to its impact on mitochondrial function; this may affect both triggering and amplification of insulin secretion. We therefore assessed insulin secretion in response to glucose, 10 mm α-ketoisocaproic acid (α-KIC), a mitochondrial fuel and 35 mm KCl, which bypasses metabolic coupling and directly depolarizes the plasma membrane. In islets from pre-diabetic β-Tfb1m−/− mice, insulin secretion in response to 16.7 mm glucose, but not 2.8 mm glucose, was significantly reduced (−1.6-fold, P < 0.05) compared with control islets (Fig. 5F). Moreover, insulin secretion in response to α-KIC and KCl was also impaired in β-Tfb1m−/− islets (−2.2-fold, P < 0.05 for α-KIC; −3.9-fold, P < 0.001 for KCl; Fig. 5F). The reduced KCl response could indicate a loss of insulin from β-Tfb1m−/− islets or an unspecific perturbation of the exocytotic machinery. Indeed, islet insulin content was reduced from 10.1 ± 0.8 ng/islet in control islets to 5.5 ± 0.8 ng/islet in β-Tfb1m−/− islets (n = 4, P = 0.006). However, mRNA expression of Ins1 and Ins2 genes were unchanged in β-Tfb1m−/− islets (Supplementary Material, Fig. S1). Thus, these results indicate that disruption of Tfb1 m in β-cells impaired fuel-stimulated secretion and biosynthesis of insulin.
Given the reduction in insulin content, which can be attributed to either impaired production or increased turnover of insulin granules, we examined the volume density of insulin granules (Nv; corresponds to the total number of insulin granules) prior to the development of diabetes. Indeed, Nv was reduced from 10.7 ± 0.7 granules/µm3 in control β-cells to 4.7 ± 1.1 granules/µm3 in Tfb1m-deficient β-cells (Fig. 5G; n = 3, P < 0.01), reflecting the reduction in insulin content. Moreover, the surface density of insulin granules (Ns), which is a measure of docked granules, was similarly reduced from 0.64 ± 0.07 granules/µm2 in control β-cells to 0.34 ± 0.04 granules/µm2 in Tfb1m-deficient β-cells (Fig. 5H; n = 3; P < 0.05). Because insulin biosynthesis largely depends on fuel metabolism (18), mitochondrial dysfunction may underlie the reduced number of insulin granules in β-Tfb1m−/− islets and hence insulin content, functionally contributing to abrogated insulin secretion.
Perturbed islet morphology and β-cell loss in β-Tfb1m−/− mice
Next, we examined the impact of Tfb1m deficiency in β-cells on islet morphology and β-cell mass. Immunostaining of pancreatic sections with antibodies against insulin, glucagon and somatostatin revealed normal islet architecture in pre-diabetic β-Tfb1m−/− mice (Fig. 6A). Nevertheless, unbiased quantitative stereological analysis revealed small, but significant, decreases in β-cell mass and β-cell area per pancreatic area in pre-diabetic β-Tfb1m−/− mice, whereas mean islet size and islet number were unchanged (Fig. 6A and Fig. 7A–D). We also examined an intermediate group of β-Tfb1m−/− mice, which were hyperglycemic but had not yet developed severe clinical symptoms of the disease (plasma glucose: 15.5 ± 1.0 mm; age: 89 ± 8 days). Indeed, reflecting the rise in plasma glucose, islets from β-Tfb1m−/− mice at the intermediate stage exhibited additional decreases in β-cell area per pancreatic area and islet number, whereas mean islet size was not significantly changed (Fig. 6B and Fig. 7E–G). Accordingly, immunostaining of pancreatic sections from late-stage, overtly diabetic, β-Tfb1m−/− mice (plasma glucose: 26.3 ± 0.8 mm; age: 149 ± 19 days) showed dramatic reductions in β-cell mass, mean islet size, islet number and β-cell area per pancreatic area (Fig. 6C and Fig. 7H–K), whereas α-cell mass remained unchanged (Supplementary Material, Fig. S4). To further illustrate the progressive nature of the β-cell loss in β-Tfb1m−/− mice, we plotted mean β-cell area per pancreatic area of all islets examined versus animal age (Fig. 7L–M); a clear correlation of declining β-cell mass with age was observed in β-Tfb1m−/− mice (r2 = −0.43; P < 0.0001). Clearly, Tfb1m deficiency in β-cells was accompanied by progressive loss of β-cell mass.
Reduced cell viability and increased inflammation in β-Tfb1m−/− islets
To investigate the mechanisms underlying β-cell loss, we challenged islets from pre-diabetic β-Tfb1m−/− mice ex vivo with 27.8 mm glucose for 24 h. This glucose concentration was similar to the observed in vivo concentration in the β-Tfb1m−/− mice after onset of diabetes (26.3 ± 0.8 mm). This treatment resulted in a significant increase of cell death in islet cells from β-Tfb1m−/− mice (Fig. 8A and B).
To confirm these findings in vivo, we measured markers for apoptosis and/or necrosis in pancreatic tissue sections: abnormal nuclei and apoptotic bodies were counted in islets from late-stage β-Tfb1m−/− and control mice. The abundance of abnormal nuclei was >10-fold greater in β-Tfb1m−/− islets compared with control islets (P< 0.01; Fig. 8C). Moreover, 0.4% of β-Tfb1m−/ islet cells exhibited apoptotic bodies, whereas none were found in islet cells in control mice (Fig. 8D). Hence, accelerated β-cell death is likely to contribute to loss of β-cell mass in β-Tfb1m−/− mice, and subsequently glucose intolerance and development of diabetes.
We also found that 71 ± 22% of islets were surrounded by macrophages in overtly diabetic mice as opposed to 32 ± 8% of islets in control mice (P< 0.05; Fig. 8E). Moreover, 72 ± 11% of the islets in β-Tfb1m−/− mice were surrounded by CD4+ cells compared with 32 ± 4% in control mice (P< 0.05; Fig. 8F). These findings demonstrate that Tfb1m deficiency results in progressive β-cell death, which is accompanied by aberrant recruitment of proinflammatory cells, suggestive of a local inflammatory response.
ROS generation and antioxidant enzymes in β-Tfb1m−/− islets
To shed light on the mechanism of increased cell death in β-Tfb1m−/− islets, we investigated glucose-induced production of ROS. We used 2′,7′-dichlorofluorescein diacetate to measure ROS production in freshly isolated islets from pre-diabetic β-Tfb1m−/− mice during a 30-min incubation in 16.7 mm glucose. A more pronounced increase in ROS formation was observed in β-Tfb1m−/− islets compared with that in control islets (3.3-fold, P < 0.05; Fig. 8G). Treatment of islets with 16.7 mm glucose and 1 µm rotenone, which is known to additionally increase ROS production in normal pancreatic islets (19), further increased ROS formation in control islets; however, no further increase was observed in β-Tfb1m−/− islets.
After formation, ROS levels are controlled by an array of crucial antioxidant enzymes, some of which we then examined in islets from pre-diabetic β-Tfb1m−/− mice. The mRNA levels of Glrx2 in freshly isolated islets were 1.2-fold higher (P < 0.05) in β-Tfb1m−/− islets, whereas the levels of Cat and Sod2 mRNA were similar (Fig. 8H). Up-regulation of Glrx2 suggests that Tfb1m deficiency in islets may perturb cellular thiol homeostasis owing to alterations in redox status. The finding of unchanged levels of Sod2 mRNA in β-Tfb1m−/− islets compared with control islets suggests that either mitochondrial superoxide production is not increased or that these islets have reduced capacity to induce oxidative defenses.
To test susceptibility to oxidative stress, the islets from pre-diabetic β-Tfb1m−/− mice were again cultured for 24 h in 27.8 mm glucose and mRNA levels of antioxidant enzymes were quantified. Under this stressful condition, the mRNA levels of Glrx2 and Sod2 in β-Tfb1m−/− islets were actually decreased by 1.6- and 3.0-fold, respectively (P < 0.001), whereas Cat expression did not change (Fig. 8I). Thus, Tfb1m deficiency was associated with an increased vulnerability to oxidative stress whereas high glucose provoked ROS production. Taken together, these results show an impairment of stress responses, known to trigger cell death (20), in the severely respiratory chain-deficient β-Tfb1m−/− islets.
TFB1M mRNA expression and insulin secretion in human islets
We recently reported that carriers of rs950994, a single-nucleotide polymorphism in intron 2 of TFB1M, are at greater risk of developing T2D (5). We have demonstrated, in a limited set of human islets, using microarray analysis, that A-allele carriers (n = 24) exhibit a ∼20% decrease in TFB1M mRNA levels compared with non-risk GG-allele carriers (n = 26) (5). To confirm the relevance of our experimental findings in β-Tfb1m−/− mice, we extended the analysis of human islets, using RNA sequencing in islets from 129 donors. We found that the risk A-allele exerted a negative gene-dosage effect on TFB1M expression: one allele reduced expression by 17% and two alleles by 34% based on the FPKM value (Fragments Per Kilobase of transcript per Million mapped reads) (Fig. 9A). The expression of all seven exons in TFB1M was similarly reduced in islets from carriers of the risk variant (data not shown). Thus, there was no differential effect on splicing of the TFB1M transcript. Moreover, the reduction in TFB1M expression was paralleled by a similar impact on insulin secretion: the fold increase (from 2.8 to 16.7 mm glucose; 1 h) was reduced by 29 and 44%, respectively, depending on the presence of one and two A-alleles (Fig. 9B). These data support the notion that reduced expression of TFB1M conferred by the risk A-allele is associated with reduced insulin secretion in human islets. It also justifies the use of β-Tfb1m−/− mice as a model to understand the pathogenetic role of TFB1M in human T2D.
The role of mitochondrial metabolism in the regulation of insulin secretion from pancreatic β-cells is essential. An elevation of extracellular glucose is translated into a series of intracellular metabolic events in the β-cell that control the timing and extent of insulin release (9,10). It was not until recently, however, that the function of the TFB1M gene, which controls mitochondrial metabolism, was linked to the common form of T2D (5). This finding should not be confused with the pathophysiology of mitochondrial diabetes, which is a rare maternally inherited condition caused by mutations in mitochondrial DNA (10). Instead, TFB1M is a nuclear-encoded protein that controls protein translation in the mitochondria. In agreement with our previous findings (5), we here show that carriers of the TFB1M risk variant exhibited lower expression of TFB1M mRNA in islets and reduced insulin secretion. In support of these findings, conditional targeting of the Tfb1m locus in murine β-cells resulted in perturbed mitochondrial function, number and morphology, β-cell dysfunction, loss of β-cell mass and subsequent development of diabetes. Taken together, the human and mouse data highlight the possible role of this gene in the development of T2D.
We recently demonstrated a link between T2D and TFB1M by showing an association of the gene with impaired insulin secretion and increased future risk of T2D (5). Carriers of the pathogenetic variant exhibit reduced expression of TFB1M in islets, resulting in reduced levels of mitochondrial proteins, particularly those that constitute the respiratory complexes responsible for OXPHOS. Moreover, mice heterozygous for an inactive Tfb1m allele exhibit a ≈50% reduction in TFB1M protein and mitochondrial-encoded proteins in islets, (5). We also found that islet ATP production is impaired in these heterozygous mice. Consequently, Tfb1m+/− mice exhibit impaired insulin secretion; similar findings were made when Tfb1m was silenced in clonal β-cells.
In the present study, we show that upon inactivation of both Tfb1m alleles in pancreatic β-cells, the levels of mitochondrial-encoded proteins were reduced, which is most likely due to perturbed translation. This can be attributed to reduced adenine dimethylation of the 12S subunit of mitochondrial rRNA, known to result in its destabilization (12). However, the impact on nuclear-encoded proteins was more complex. The mitochondrial functional deficiency may lead to a nuclear response, attempting to devise compensatory effects, in some instances leading to increased levels of nuclear-encoded mitochondrial proteins. Conversely, it is also known that the stability of some nuclear-encoded proteins in the mitochondria may depend on critical mitochondrial-encoded proteins (21). Whether this perturbed relationship between nuclear and mitochondrial proteins contributed to the cellular dysfunction was not determined. Nevertheless, these changes could be of relevance because stochastic protein imbalances may trigger a mitochondrial unfolded protein response (22). This notwithstanding, together, these perturbations of mitochondrial protein levels most likely reduced electron transport and subsequently OXPHOS, which led to mitochondrial dysfunction in β-Tfb1m−/− islets. As a consequence, β-cell stimulus-secretion coupling was perturbed, insulin release and content reduced, and the β-Tfb1m−/− mice progressively became glucose intolerant, hyperglycemic and ultimately developed diabetes. This phenotype seems to have been brought about by a combination of β-cell dysfunction and eventually loss of β-cell mass, driven by the mitochondrial dysfunction. The temporal sequence of events indicates that cellular dysfunction preceded cell loss. This process is very similar to the development of T2D (2,3), although the velocity of events has been accelerated in our model.
Following the initial perturbation of β-cell function, owing to the mitochondrial impairments, β-cell mass was progressively reduced upon loss of Tfb1m. This could be attributed to both apoptosis and necrosis. Indeed, there are several murine and experimental models, in which deficiencies in electron transport and/or OXPHOS result in extensive cell death (23,24), most likely due to energy deficiency. We believe that a similar process promoted cell death in β-Tfb1m−/− islets. Under physiological conditions, functional β-cells are metabolically very active cells; they rely foremost on ATP produced in mitochondria rather than through glycolysis. In a pathological situation, dramatic reduction of ATP production could lead to death of any cell (25), and depletion of ATP can transform an ongoing apoptotic process into necrosis (26). The energy depletion in β-Tfb1m−/− β-cells owing to electron transport dysfunction and impaired OXPHOS could by itself account for cell death.
It has also become increasingly clear that pancreatic β-cells are characterized by a weak antioxidant defense (27). Only limited amounts of ROS are formed in mitochondria because O2 is reduced to H2O. Experimental studies have demonstrated that a global impairment of mitochondrial DNA expression in the mouse does not lead to increased ROS production, although apoptotic cell death is increased (23,28,29). Here, we challenged β-Tfb1m−/− islets with glucose and observed a marked increase of ROS production. Although this process is attractive as a mechanism for cell death in β-Tfb1m−/− islets, the result must be interpreted with caution as the subcellular origin of this ROS production is unknown. Several other enzymes in metabolism, besides the respiratory chain, are known to produce ROS, e.g. α-ketoglutarate dehydrogenase in the TCA-cycle (30) as well as intrinsic NADPH oxidase activity in β-cells (31). The macrophages recruited to the β-Tfb1m−/− islets, perhaps to clear apoptotic cells, may also be a source for extrinsic production of ROS. Regardless of the source, the observed increase of ROS production in response to elevated glucose in β-Tfb1m−/− islets may contribute to the observed cell death. We found that β-Tfb1m−/− islets lacked the capacity to induce mRNA expression of scavenging enzymes, e.g. Sod2, in response to oxidative stress. In fact, mRNA levels of Glrx2 and Sod2 in β-Tfb1m−/− islets were reduced. These findings indicate that the bioenergetic deficiency in β-Tfb1m−/− islets led to a decrease of cellular stress responses. This is likely to increase the propensity for cell death. In addition, a perturbed balance between nuclear and mitochondrial proteins may trigger a mitochondrial unfolded protein response. This may further increase the risk of cell death.
Our findings of cell death may also be relevant for human T2D, where β-cell loss and apoptosis in islets have been reported (32). The possibility of necrosis with attendant inflammation was confirmed by an accumulation of macrophages and CD4+ cells in the vicinity of and throughout the β-Tfb1m−/− islets. This finding is noteworthy, given the recent interest in local islet inflammation in the pathogenesis of T2D (29). Taken together, our data suggest that the disrupted translation of mitochondrial proteins caused a combined insult of a bioenergetic deficiency, which depleted ATP, and increased ROS production in β-Tfb1m−/− β-cells. It is not unlikely that the bioenergetic deficiency and increase in ROS production were mechanistically linked because impaired electron transport is known to increase ROS formation (23,28,29). Nevertheless, such an insult may lead to β-cell death via apoptosis and necrosis.
Previous studies of islets from humans with T2D have shown that mitochondrial membrane polarization and the ATP/ADP ratio are unresponsive to glucose (33). Moreover, mitochondrial density is increased and mitochondrial morphology altered in islets from T2D patients (4). In mice, a β-cell-specific knockout of Tfam, the major regulator of expression of mitochondrial-encoded genes, results in a phenotype similar to that of the β-Tfb1m−/− mice: insulinopenic diabetes evolves owing to a combination of β-cell dysfunction and loss of β-cell mass (34). However, no signs of apoptosis could be detected in Tfam null mice, but in vitro challenges of islets were not performed. These findings, however, are clearly reminiscent of what we have observed here and in our previous studies of TFB1M (5). An important difference is that the previous observations of mitochondrial dysfunction in islets from humans with T2D do not prove a causal relationship (4); they may be secondary to the pathological metabolic milieu. Furthermore, whereas Tfam is critical for the transcriptional regulation in mitochondria, it has not been genetically or causally linked to human T2D, whereas this link has been shown for TFB1M (5).
Taken together, our findings point to a pathogenetic and clinically important role of TFB1M in T2D. Our data propose a molecular mechanism for TFB1M deficiency in β-cells and its involvement in the pathogenesis of T2D: mitochondrial dysfunction paves the way for the development of diabetes owing to combined β-cell dysfunction and loss of β-cell mass (Fig. 9C).
MATERIALS AND METHODS
Generation of mice with a β-cell-specific knockout of Tfb1m
To generate a conditional knockout of Tfb1m in β-cells, we used Tfb1mloxP/loxPmice in which exon 3 of the Tfb1m locus was flanked by two loxP sites (12). Heterozygous Tfb1m+/loxP mice were mated to heterozygous transgenic mice expressing Cre recombinase under the control of a rat insulin 2 gene promoter (Rip-Cre) (35). Mice with the genotype Tfb1m+/loxP,+/Rip−Cre were recovered from this cross and bred with Tfb1mloxP/loxPmice to generate Tfb1mloxP/loxP,+/Rip−Cre (β-Tfb1m−/−; β-cell-specific Tfb1m knockout mice) and Tfb1mloxP/loxP (β-Tfb1m+/+; control) mice. Both male and female animals were used for the experiments. The floxed Tfb1m+/− mice used in this study have been backcrossed more than nine times onto the C57BL/6J background. For the experiments, mice were defined as early stage/pre-diabetic (control: glucose 9.6 ± 0.4 mm, age: 73 ± 20 days; β-Tfb1m−/−: glucose 9.4 ± 0.3 mm, age: 76 ± 17 days), intermediate stage (control: glucose 8.1 ± 0.4 mm, age: 85 ± 7 days; β-Tfb1m−/−: glucose 15.5 ± 1.0 mm, age: 89 ± 8 days) and late stage/diabetic (control: glucose 8.4 ± 0.1 mm, age: 138 ± 19 days; β-Tfb1m−/−: glucose 26.3 ± 0.8 mm; age: 149 ± 19 days); ages are given as mean ± SD. Mouse breeding and handling were carried out according to procedures approved by the regional animal ethical committee in Lund, Sweden.
Mouse and human islets
Pancreata were removed from mice, chopped and digested at 37°C in 3 ml of Hanks' balanced salt solution containing 0.8 mg/ml of collagenase for 7–8 min with continuous shaking. Islets were hand-picked under a stereo microscope and incubated at 37°C overnight in a humidified atmosphere of air and 5% CO2 in RPMI-1640 medium (11.1 mm glucose) supplemented with 10% fetal calf serum, 100 U/ml penicillin and 100 μg/ml streptomycin sulfate.
Human islets were from deceased donors obtained from the Nordic Center for Clinical Islet Transplantation by courtesy of Professor Olle Korsgren, Uppsala University, Sweden. Islets were processed at the Human Tissue Laboratory at Lund University Diabetes Centre. Purity varied from 23 to 90%, as determined by supravital dithizone staining as described (36). The islets were cultured in CMRL 1066 (ICN Biomedicals, Costa Mesa, CA, USA) supplemented with 10 mmol/l HEPES, 2 mmol/l l-glutamine, 50 μg/ml gentamicin, 0.25 μg/ml Fungizone (GIBCO, BRL, Gaithersburg, MD, USA), 20 μg/ml ciprofloxacin (Bayer Healthcare, Leverkusen, Germany) and 10 mmol/l nicotinamide at 37°C in a humidified atmosphere of 95% air and 5% CO2 for 1–9 days prior to analysis. All islet donors had given consent to donate organs for medical research. The procedures were approved by the human ethical committees in Uppsala and Lund.
The genotype of the mice was verified by PCR analysis of DNA isolated from ear clip DNA (DNeasy Blood and Tissue kit; QIAGEN). The presence of the Rip-Cre gene was confirmed by the amplification of an approximately 400-bp product, using forward 5′ GCATTACCGGCTGATGCAACGAGTGATGAG and reverse 5′ GAGTGAACGAACCTGGTCGAAATCAGTGCG primers. PCR conditions were 95°C for 1 min, followed by 35 cycles of 95°C for 30 s, 68°C for 3 min and 68°C for 3 min. Genomic PCR for detection of the lox p allele was performed using forward 5′ ATGTTTACAGGCTTAGTTGAA and reverse 5′ TAGTAGAGAATACTGCCACAG primers. PCR products were 198 bp for the wild-type allele and 244 bp for the lox p allele. PCR amplification was performed at 95°C for 5 min, and 35 cycles of 95°C for 30 s, 55°C for 30 s, 72 °C for 30 s and the final extension at 72°C for 5 min.
The genotyping of the human islets was performed on the Illumina HumanOmniExpress 12v1 chips (Illumina, Inc., San Diego, CA, USA) according to manufacturer's instruction. Genotypes for rs950994 were called using Illumina Genome studio software.
RNA extraction, reverse transcription and quantitative real-time PCR
Total RNA was extracted from islets, using RNAeasy RNA purification kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer's instructions. RNA concentration was determined using a NanoDrop Spectrophotometer (Thermo Scientific) and its quality assessed with the Agilent Bioanalyzer. Equal quantities of total RNA were reverse-transcribed using RevertAid™ First-Strand cDNA synthesis kit (Fermentas, Vilnius, Lithuania) in reactions containing 500 ng of total RNA, 50 mm Tris–HCl (pH 8.3), 1 mm dNTPs, 200 ng of random hexamer primers, 50 mm KCl, 4 mm MgCl2, 10 mm dithiothreitol, 200 U of RevertAid™Moloney murine leukemia virus reverse transcriptase and 20 U of RiboLock™ ribonuclease inhibitor in a final volume of 20 μl. The reaction cycle consisted of 10 min of incubation at 25°C followed by 60 min of incubation at 42°C and termination of the reaction by heating at 70°C for 5 min. Samples were stored at −20°C.
Q-PCR was performed on an ABI Prism 7900 HT system (Applied Biosystems), using TaqMan chemistry. All reactions were performed in triplicate in a final volume of 20 μl containing 5 μl of TaqMan Universal PCR Master Mix, 0.25 μl of a corresponding TaqMan assay and 12 ng of cDNA template. cDNA was amplified by 40 cycles of 95°C for 15 s and 60°C for 1 min. mRNA levels of genes were quantified using predesigned TaqMan Gene Expression assays (Applied Biosystems, Supplementary Material, Table S2). Gene expression was determined by the absolute quantification method and normalized to mRNA levels of Hprt or Vdac1.
RNA-sequencing libraries were generated from 129 islet donors (TruSeq RNA sample preparation kit, Illumina) and sequenced on an Illumina HiSeq 2000 using paired-end chemistry and 100-bp cycles to an average depth of 32 M read pairs/sample. Reads were aligned to hg19, using STAR (37), and read count calculated by HTSeq-count (http://www-huber.embl.de/users/anders/HTSeq/doc/count.html). Data were log2-transformed and normalized, using trimmed mean of M-values as implemented in the R-package edgeR. The global genomic and transcriptomic analyses of our human islet depository are under review elsewhere (Fadista et al. submitted). The full data set will be deposited online for public use once the study has been published.
Islets were solubilized in homogenization buffer (100 mm HEPES, 9 m urea, 1% Triton X-100, 2 mm EDTA, pH 7.2) mixed with 1:100 (v/v) protease inhibitor cocktail (Sigma, St Louis, MO, USA). Protein concentration was determined by the BCA method, protein lysates mixed 1 : 5 with loading buffer (100 mm HEPES, 10% SDS, 10% dithiothreitol, 20% glycerol, pH 7.2), and 40 µg of total protein were run onto a 12% SDS–PAGE gel and subsequently blotted onto PVDF membranes. Proteins for TFB1M, ND1 (complex I), NDUFB8 (complex I), SDHB (complex II), COX1 (complex IV) and ATP5A (FOF1 ATP synthase, subunit α) were detected as described previously (5). β-Tubulin was detected with primary rabbit polyclonal antibody (Abcam, Cambridge, UK) in dilution of 1 : 500. Horseradish peroxidase-coupled goat-anti-rabbit IgG (1:6000) and goat-anti-mouse IgG (1 : 6000) (Santa Cruz Biotechnology, Santa Cruz, CA, USA) were used as a secondary antibody. Blots were developed with enhanced chemiluminescence (Amersham Biosciences), and detection was performed using Molecular Imager® ChemiDoc™ XRS with Image Lab software (BioRad). For loading, control blots were stripped with Restore Western Blot Stripping Buffer (Thermoscientific).
Primer extension analysis
A set of two primers (MM49 and MM13) was used for primer extension. MM49 (12 bases, 5′ GTGTAATTTTAC 3′) is complementary to a region close to the 5′ end of 12S rRNA, and its extension product (P1; 46 bases) was used as a 12S rRNA loading control. MM13 (12 bases, 5′ ATTATTCCAAGC 3′) is complementary to a region positioned downstream of the dimethylated A1006 and A1007. Extension of this primer is inhibited by the presence of methylation and leads to accumulation of an extension product P2 (17 bases). Primers were labeled with γ-32P-ATP by using T4 polynucleotide kinase. Labeled primers were purified using MicroSpin G-25 chromatography columns (GE Healthcare) and annealed to RNA from islets of control and β-Tfb1m−/− mice. Reactions were incubated at 65°C for 5 min and placed on ice. M-MuLV Reverse Transcriptase (Stratagene) was added to 5 U/μl together with dNTPs (0.25 mm/μl each). The reactions were incubated for 1 h at 37°C. Extension products were precipitated and resuspended in loading buffer (formamide, 1 mg/ml bromophenol blue, 1 mg/ml xylene cyanol, 10 mm EDTA; pH 8.0). Primer extension products were resolved using denaturing urea-PAGE and detected by autoradiography.
Blood glucose concentrations were repeatedly measured (bi-weekly) in non-fasted control and β-Tfb1m−/− mice (n = 5–8). The blood glucose concentration was determined using a FreeStyle Lite Glucose Analyser (Abbot Diabetes Care, Inc., USA).
Intraperitoneal glucose tolerance test
IPGTTs were performed in fasted mice; food was removed 5 h before the glucose challenge. The mice were anesthetized as previously described (38). Blood was collected retroorbitally at 0, 10, 30, 60 and 120 min after intraperitoneal injection of glucose (2 g/kg body weight). The plasma glucose concentration was measured by the FreeStyle Lite Glucose Analyser (Abbott, Abbott Park, Ill, USA). Insulin concentration was assayed by the Ultrasensitive Mercodia insulin enzyme-linked immunosorbent assay (Mercodia, Uppsala, Sweden).
Insulin secretion assay
Islets were preincubated in Krebs-Ringer bicarbonate buffer (KRBB) containing (mm) 114 NaCl, 4.7 KCl, 1.2 KH2PO4, 1.16 MgSO4, 20 HEPES, 25.5 NaHCO3, 2.5 CaCl2 at pH 7.2 with 2 mg/ml BSA and 2.8 mm glucose for 1 h at 37°C. Next, groups of three islets were incubated at 37°C for 1 h in 300 μl KRBB containing 2.8 mm glucose, 16.7 mm glucose, 2.8 mm glucose and 10 mm α-KIC, or 16.7 mm glucose and 35 mm KCl. The buffer was removed, and secreted insulin was determined by the Mercodia insulin enzyme-linked immunosorbent assay.
Islet insulin content
Islets were washed in PBS, mixed with ethanol/HCl at 4°C and sonicated. Thereafter, islets were repeatedly frozen and thawed and centrifuged for 5 min at 16 000 g. Supernatants were stored at −20°C until assayed by the Mercodia insulin enzyme-linked immunosorbent assay (Uppsala, Sweden).
Mitochondrial membrane potential
Islets were loaded with 400 nm TMRM for 2 h in KRBB containing 2.8 mm glucose, thus permitting analysis in “quench mode” (39). The chambered cover glass was inserted into a temperature-controlled (37°C) and CO2-controlled (5%) incubation chamber on the stage of a Zeiss LSM510 inverted confocal fluorescence microscope. TMRM was excited at 543 nm, and emission was detected with a 585-nm long-pass filter. Islets were stimulated with 16.7 mm glucose, oligomycin and FCCP at the times indicated in Figure 3A to investigate changes in mitochondrial membrane potential. Fluorescence intensity was measured in nine islets from control mice (n = 5) and 26 islets from β-Tfb1m−/− mice (n = 7; age: 76 ± 17 days).
OCR was measured in islets using the XF (extracellular flux) analyzer XF24 (Seahorse Bioscience, Billerica, MA, USA), as previously described in detail (40). Following a pre-incubation at 2.8 mm glucose for 1 h, the OCR was assayed at 2.8 mm glucose subsequent to a transition to 16.7 mm glucose. AUC was calculated under the high glucose-stimulated condition.
Islets were washed in PBS, lysed in Tris–EDTA buffer and ATP measured with a luciferase-based luminescent assay according to the manufacturer's instructions (BioThema, Handen, Sweden). The rate of mitochondrial ATP production in digitonin (400 μg/ml)-permeabilized islets was measured as previously described (41) with a mixture of 15 mm glutamate and 15 mm succinate as stimulating substrate.
Pancreata were dissected out and weighed before they were fixed overnight at 4°C in Stefanini's solution (2% paraformaldehyde and 0.2% picric acid in 0.1 M PBS; pH 7.2). The tissue was washed thoroughly 3 × 24 h at 4°C, in Tyrode's solution containing 10% sucrose, and thereafter frozen on dry ice. Sections (10 μm) were cut and thaw-mounted on slides.
Antibodies were diluted in PBS, pH 7.2, containing 0.25% BSA and 0.25% Triton X-100. Sections were incubated with primary antibodies against insulin (dilution: 1 : 5000, code: 9003, EuroDiagnostica, Malmö, Sweden), glucagon (dilution: 1 : 5000, code: 7811, EuroDiagnostica), somatostatin (dilution: 1 : 1600, code: sc-7819, Santa Cruz Biotechnology, California, USA), macrophages (dilution: 1 : 1000, T-2007, BMA Biomedicals, Augst, Switzerland), chromogranin A and B (dilution: 1 : 25, code: CAB, EuroDiagnostica) or CD4 (dilution: 1 : 100, code: ab64145, AbCam) overnight at 4°C in moisture chambers. Sections were rinsed 2 × 10 min in PBS with 0.25% Triton X-100 and incubated 1 h at RT with secondary antibody with specificity for rabbit-, guinea pig-, rat- or goat-IgG coupled to either Cy2, Texas red or AMCA (all from Jackson, West Grove, PA USA). Sections were rinsed as before and mounted in PBS/glycerol (1 : 1). For double and triple staining, the primary and the secondary antibodies were applied simultaneously.
Immunofluorescence was examined in an epifluorescence microscope (Olympus BX60, Olympus, Tokyo, Japan). Images were acquired with a digital camera (Nikon DS- 2Mv, Nikon, Tokyo, Japan). For β-cell mass quantification, all islets in randomly selected parts of each pancreas (n = 9 or more) were analyzed using NIS-Elements software (NIS-Elements 3.1, Nikon). First, total insulin-stained area and total section area were analyzed and calculated. The ratio achieved was multiplied with the pancreas weight to calculate the β-cell mass. An average of 40 islets per animal was analyzed. α-cell mass was quantified in a similar fashion in pancreatic sections immunostained for glucagon. The presence of leukocytes within, or in the vicinity of, the islets was detected by immunofluorescence staining against a subpopulation of leukocytes and chromogranin A and B to visualize the islets. One or more specific leukocyte in an islet or within 150 μm from it was identified as a positive finding. The investigator was unaware of the identity of the specimens during the analysis.
Groups of 40–50 isolated islets were fixed in Millonig buffer (2.26% NaH2PO4 and 2.52% NaON) containing 2.5% glutaraldehyde. Following washing in Millonig buffer, islets were post-fixed in 1.0% osmium tetroxide, dehydrated and embedded in AGAR 100 (Oxfors Instruments Nordiska AB, Sweden). The embedded islets were cut into 70- to 90-nm-thick ultrathin sections and put on Cu grids and contrasted with uranyl acetate and lead citrate before being examined in JEM 1230 electron microscope (JEOL-USA, Inc., MA, USA). Micrographs were analyzed with respect to mitochondrial area, granule distribution and number of docked granules. The mitochondrial area and the granule diameter were analyzed using ImageJ (NIH, freeware; see Supplementary Material, Fig. S2, for mitochondrial area). The estimated 3D diameter of the granules was calculated as described previously (42). The volume granule density (Nv) and surface density (Ns) were calculated using an in-house Matlab program.
Detection of apoptosis and necrosis
Batches of 35 islets were incubated overnight in complete RPMI-1640 medium containing 27.8 mm glucose. At the end of the incubation, apoptosis and necrosis were assessed in islets by a GFP-Certified Apoptosis/Necrosis Detection Kit (Enzo Life Sciences, Inc., USA), using a microplate analyzer (TECAN Infinite M200). In the assay, an Annexin V-EnzoGold (enhanced Cyanine-3) conjugate enables detection of apoptosis at 550/570 nm excitation/emission; the necrosis detection reagent (red) identifies late markers of apoptosis and necrosis at 546/647 nm excitation/emission.
Pancreatic sections from five control mice (age: 138 ± 19 days) and five β-Tfb1m−/− mice (age: 149 ± 19 days; ages are given as mean ± SD) were stained with DAPI to visualize nuclei. A total of 450–900 islet cells in each mouse were manually counted to determine the percentage of abnormal nuclei (defined as previously detailed (43) and apoptotic bodies.
All calculations were performed in GraphPad Prism 4.03 (GraphPad, La Jolla, California, USA) software. P-values were considered significant at <0.05. All data are presented as mean ± SEM of the indicated number of experiments or animals with the exception of mouse ages and background data on islet donors, which were given as mean ± SD. Data were analyzed using one-way ANOVA, two-way ANOVA, repeated measure of two-way ANOVA or two-tailed unpaired Student's t-test. Statistical significance of differences in total mitochondrial area was analyzed using Student's unpaired t-test and non-parametric boxplot descriptive statistics using R. In the latter non-overlapping box notches are strong indicators that medians of the two sets are different (44).
V.V.S. carried out the majority of in vivo and vitro experiments and drafted parts of the manuscript. M.A., L.E. and I.G.M. performed the morphological and morphometric studies. J.S., J.F. and P.S. performed bioinformatics analyses of RNA sequencing. L.M.N., P.S., S.M. and J.A.S. performed in vivo and vitro experiments. M.D.M. and N.-G.L. performed the methylation experiments. L.M.N. N.W., N.-G.L., L.E. and P.S. provided feedback on data interpretation and assisted in writing the manuscript. H.M. conceived the study, interpreted results and wrote the manuscript.
Support was received from these funding bodies: European Foundation for the Study of Diabetes (EFSD), Crafoordska, Knut and Alice Wallenberg, Lars Hiertas Minne, Söderberg, O.E. och Edla Johansson, and Albert Påhlsson foundations, and The Swedish Research Council.
Rita Wallén is acknowledged for preparing the EM samples. L.E. is a senior researcher at the Swedish Research Council. Laila Jacobsson is acknowledged for genotyping the mice.
Conflict of Interest statement. None declared.