ABSTRACT
In the progression of diabetes, pancreatic islet β-cells respond to increased metabolic demand with functional compensation, followed by pathogenic decompensation of mitochondria-dependent insulin secretion. It is not clear what mechanisms drive, or control, mitochondrial decompensation. Here, we report that anti-apoptotic Bcl-xL maintains mitochondrial integrity in β-cells under non-apoptotic levels of glucose stress. Prolonged glucose excess causes transcriptional reprogramming of glycolysis and β-cell identity genes, while sensitizing glucose-stimulated Ca2+ signaling and insulin secretion. Deletion of Bcl-xL amplifies this insulin hypersecretion and increases mitochondrial fusion, mitochondrial volume, and oxygen consumption, whereas ATP-coupled respiration and mitochondrial hyperpolarization become impaired. Of note, Bcl-xL-deficient β-cells have impaired Pgc-1α expression, and develop specific defects in the expression of Tfam, mitochondrial ribosomal genes, and OXPHOS components under glucose stress. Bcl-xL limits high glucose-induced mitochondrial ROS (mitoROS) levels and pharmacological normalization of mitoROS in Bcl-xL KO cells rescues glucose-induced defects in mitochondrial gene expression and changes to β-cell identity. Our data identify mitoROS as a primary retrograde driver of transcriptional re-wiring in β-cells exposed to excess glucose, and reveal Bcl-xL as an important safeguard against transcriptional and functional decompensation of β-cell mitochondria. Bcl-xL and mitoROS may thus be viable targets to prevent early β-cell dysfunction and the progression of diabetes.
INTRODUCTION
Nutrient-induced β-cell insulin secretion is essential to maintain normoglycemia, and β-cell failure is a defining event in the development of type 2 diabetes (T2D). In response to worsening insulin resistance and nutrient excess β-cells respond by increasing their total insulin output, which may compensate sufficiently to maintain euglycemia and prevent the progression toward prediabetes and diabetes. Evidence suggests that β-cell compensation involves both an initial augmentation of cellular function1,2, and a slower expansion of total β-cell mass2. However, in individuals that progress to develop diabetes, the compensatory response is transient due to a progressive development of β-cell dysfunction, and ultimately also a loss of β-cells by dedifferentiation, trans-differentiation and apoptosis.
β-Cell mitochondrial metabolism generates ATP and other metabolic coupling factors that are essential for glucose-stimulated insulin secretion3. Severe disruptions to both mitochondrial function and structure are seen in β-cells from mice and humans with T2D, indicating that mitochondrial dysfunction plays a role in the failure of nutrient-induced insulin secretion4–6. Recent findings have shown that relatively modest increases in blood glucose are associated with changes in the expression of genes related to β-cell identity and metabolism7, and that the progression of diabetes and hyperglycemia is associated with increasing disruption of transcripts and proteins involved in oxidative phosphorylation and other aspects of mitochondrial physiology8. Comparatively little is known about the mechanisms that mediate the earlier stages of functional adaptation, but augmented mitochondrial metabolism likely plays a central role9. Therefore, glucose-induced changes to β-cell mitochondrial function appear central to the development of T2D. Despite these recent insights, it remains unclear what mechanisms drive the progression from mitochondrial adaptation to failure under conditions of chronic glucose excess.
Apoptosis-regulating proteins in the Bcl-2 family have emerged as important regulators of physiological processes in non-transformed cells, including notable roles of both pro- and anti-apoptotic proteins in the regulation of cellular metabolism10. We, and others, have linked Bcl-2 family proteins to glucose-stimulated mitochondrial function and insulin secretion in pancreatic β-cells11–14. Our previous work demonstrated that anti-apoptotic Bcl-xL dampens the responsiveness of β-cell mitochondria to glucose stimulation12, and we hypothesized that these non-canonical metabolic functions could play an important role in maintaining mitochondrial integrity in β-cells during increases in metabolic demand.
In this paper, we tested this possibility by examining the relationship between β-cell gene expression, function, and mitochondrial homeostasis in Bcl-xL wild-type and knockout β-cells exposed to chronic high levels of glucose. Our results reveal a previously unrecognized role of Bcl-xL in limiting the progression of mitochondrial decompensation in β-cells under non-apoptotic levels of metabolic stress.
METHODS
Reagents
Tamoxifen (#T5648), Collagenase type XI (#C7657), Tetramethylrhodamine Ethyl Ester (TMRE, #87917), D-glucose (#G7528), Diazoxide (DM, #D9035), MitoTEMPO (#SML0737), FCCP (#C2920), Oligomycin (#O4876), Antimycin-A (#A8674) and Rotenone (#R8875) were purchased from Sigma-Aldrich (St. Louis, MO). Fura-2 AM (#F1221), MitoTracker Green FM (MTG, #M7514), Hoechst 33342 (#H3570), MitoSOX (#M36008), RPMI 1640 (#11879), Dulbecco’s Modified Eagle’s Medium (DMEM, #11995), Fetal Bovine Serum (FBS, #10438), Trypsin-EDTA (#25300), Penicillin-Streptomycin 10,000 U/ml (#15140), and HBSS (#14185) were from Life Technologies/Thermo Fisher Scientific (Carlsbad, CA). Dimethyl sulfoxide (DMSO, #BP231) was purchased from Fisher Scientific (Waltham, MA) and Minimum Essential Media (MEM, #15-015-CV) was from Corning (Corning, NY). Seahorse XFe24 Islet Capture FluxPak (103518-100) were purchased from Agilent Technologies Canada (Mississauga, ON).
Animals
Mice with tamoxifen (TM)-inducible β-cell-selective deletion of Bcl-xL were established and bred as previously described12. Pdx1-CreERTM:Bclxfl/fl mice were injected 4 consecutive days with 3 mg/40g TM injection to induce gene deletion and generate Bcl-xL knockout (BclxβKO) mice. TM-injected littermate Bclxfl/fl (BclxβWT) mice were used as controls to account for possible metabolic effects of tamoxifen15,16. To avoid potential long-term compensation for Bcl-xL deletion, islets were isolated 5-7 days after the last TM injection. Experiments were done using 14-16 week old male mice. All animal procedures were done according to national and international guidelines and approved by the University of British Columbia Animal Care Committee.
Islet Isolation and Cell Culture
Pancreatic islets were isolated by our previously described methods of collagenase digestion and filtration-based purification11. The isolated islets were hand-picked and allowed to rest overnight before further analysis or experimental cultures. Aliquots of islets from all experimental mice were collected for qPCR confirmation of tamoxifen-induced gene deletion. For single cell microscopy the islets were dispersed and seeded on 25 mm glass coverslips17. The effects of nutrient excess were examined by culturing intact islets or dispersed islet-cells for 6 days in RPMI completed with 10% FBS and 2% Penicillin-Streptomycin and containing either 11 mM glucose (normal glucose; NG) or 25 mM glucose (high glucose; HG).
Respiratory Control Analysis
Islet respiration was quantified using the Seahorse XFe24 Analyzer (Agilent Technologies). Intact islets were washed with PBS twice and loaded into XFe24 islet capture microplates in Seahorse XF RPMI Medium pH 7.4 (#103576) supplemented with 3 mM glucose, 2 mM Glutamine and 2 mM Sodium Pyruvate and incubated at 37°C in a non-CO2 incubator for 60 min before loading into the XFe24 Analyzer. After the basal oxygen consumption rate (OCR) measurement reached a steady state, the wells were injected with either – 1) increasing concentrations of glucose followed by 1 µM oligomycin; or 2) exposed to a mitochondrial stress test, consisting of the sequential injection of 15 mM glucose, 1 µM oligomycin, 2 µM FCCP and finally a combination of 1 µM rotenone and 1 µM antimycin A. The amount of ATP-coupled respiration, ATP coupling efficiency, proton leak and spare respiratory capacity were calculated as described in18. OCR values were normalized to total DNA per well quantified on QubitTM Flurometer (Q32857 Invitrogen) using QubitTM dsDNA HS Assay Kit (Q32854).
Ca2+ Imaging
Intact islets were pre-cultured 2-6 days on glass coverslips in RPMI containing glucose concentrations as indicated. Cytosolic Ca2+ signaling was then measured using Fura-2AM fluorescence microscopy, essentially as previously11. Briefly, islets were loaded with 5 µM Fura-2 for 30 min in RPMI containing 3 mM glucose and transferred to a 2 ml imaging chamber for continuous perifusion at 2.5 ml/min with Ringer’s solution containing 5.5 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 20 mM HEPES, and 144 mM NaCl and varying glucose concentrations. Before image acquisition, the perifused islets were equilibrated for 30 min with 3 mM glucose Ringer’s and Ca2+ changes then recorded in response to glucose and other stimuli, as indicated. Cytosolic Ca2+ levels are expressed as the Fura-2 fluorescence intensity ratio (F340/F380).
Confocal Imaging and Analysis of Mitochondria
Live islet-cells were imaged in a Tokai Hit INUBTFP-WSKM stage-top incubator at 37°C and 5% CO2 on a Leica SP8 Laser Scanning Confocal Microscope (Concord, Ontario, Canada). Optimized 2D and 3D confocal image acquisition, as well as quantitative extraction of mitochondrial features, was done according to our recent pipeline for mitochondrial analysis19. For automated batch analysis we used the associated Mitochondria Analyzer plugin that we developed for the ImageJ/Fiji and have made available online20,21.
Machine Learning-based Classification of Mitochondrial Morphology and Networking
For machine learning-based classification of 3D mitochondrial features we first generated a training set based on image stacks of MTG-labelled mouse islet-cells and MIN6 cells expressing mitochondria-targeted YFP. From these images, morphological and networking descriptors were quantified for a total of 2190 mitochondria and the data was grouped into 7 clusters by unsupervised K-means cluster analysis in the XLSTAT software (Addinsoft, NY). The clusters were visually examined and merged based on similarity until 4 major groupings remained, which we classified as Punctate, Tubular, Filamentous, and Highly Complex based on their visual characteristics. The principal features of each morphological classification are summarized in Table 1. All 2190 mitochondria were then manually inspected to verify their correct classification and reassignments were made as necessary. The final training set was then produced by re-quantifying the morphological and networking descriptors of all mitochondria within each of the classifications.
After establishing the training set, it was used to classify mitochondria of experimental cells using a Random Forest Classification algorithm in XLSTAT. The mitochondrial profile of a cell was then established as the proportion of each mitochondrial “morpho-types” normalized to the total mitochondrial volume in the cell. A summary of the workflow is shown in Supplemental Fig. 1.
Mitochondrial ROS Imaging
Dispersed islet cells were pre-cultured for 6 days on glass coverslips in either NG or HG RPMI with or without addition of 500 nM of the mitochondria-targeted superoxide scavenger MitoTEMPO. Mitochondrial ROS was then measured using MitoSOX Red superoxide indicator. Briefly, cells were washed and then stained with 5 µM MitoSOX in phenol red free RPMI protected from light and incubated at 37°C and 5% CO2 for 15 minutes before being imaged using a Leica DMI6000 inverted microscope equipped with an HCX Plan FLUOTAR L 20x objective, a DFC365FX digital camera, 546/12x excitation and 605/75m emission filters.
Islet Insulin Secretion and Insulin Content
Insulin secretion was measured under static incubations from 15 size-matched islets. Following 6-day pre-culture, the islets were collected in 1.5 ml low retention microcentrifuge tubes and equilibrated for 1 hour in Kreb’s Ringer’s buffer (KRB; 129 mM NaCl, 4.8 mM KCl, 1.2 mM MgSO4, 1.2 mM KH2PO4, 2.5 mM CaCl2, 5 mM NaHCO3, 10 mM HEPES and 0.5% BSA) with 3 mM glucose. The same groups of islets were then transferred between tubes containing 500 µl of KRB with glucose and treatments as indicated, moving sequentially from low to high glucose conditions. After each 60 min incubation step the islets were spun down at 4°C, supernatant was collected for measurement of secreted insulin, and the islets were re-counted before transfer to the next tube. Islet insulin content was extracted by washing twice with chilled PBS and freeze-thawing in 100 μl of RIPA buffer at -20°C before assay. Insulin concentrations were measured using the Ultrasensitive Insulin ELISA kit (Alpco #80-INSMSU-E10) according to manufacturer’s manual. Luminescence was measured at 450 nm on a SpectraMaxL luminometer (Molecular Devices)
RT-qPCR
For each sample, 40 to 50 islets were preserved in 40 μl RNAlater (Qiagen #76104) at -80°C before RNA isolation. Total RNA was isolated using the RNEasy Mini Kit (Qiagen #74106) and measured by NanoDropTM 2000 (Thermo Fisher, Carlsbad, California). Freshly isolated RNA (100 ng/sample) was reverse transcribed using qScript cDNA synthesis kit (Quanta Bioscience #95047-500) and cDNA was stored at -20°C until use. The qPCR reactions were run using: 4 μl PerfeCTa SYBR green Fastmix (Quanta Biosciences #95072), 5 μM forward and reverse mixed primers, 1 μl cDNA, and 1.5 μl ddH2O. Primer efficiencies were validated by serial dilutions and amplicon size confirmed by agarose gel electrophoresis. Validated primer sequences are listed in Table 2. Expression levels were assayed in triplicates on 384-well microplates (Life technologies #4309849) on a ViiA7 Real-Time PCR machine (Applied Biosystems, Foster City, California) and normalized to the Actb housekeeping gene.
RNA-Sequencing
Sample preparation and sequencing
RNA was extracted from whole pancreatic islets using Trizol Reagent (Life Technologies, 15596026) and ethanol precipitation with glycogen as a carrier, followed by DNase-treatment using Turbo DNA Free Kit (Life Technologies, AM1907). Total RNA content was estimated using Nanodrop, and samples were prepared for sequencing from 400 ng of total RNA using NEBNext Ultra II Directional RNA Library Kit for Illumina with poy(A) mRNA enrichment (NEB, E7760, E7490) according to the manufacturer’s protocols. Libraries were quantified using Qubit fluorometer and Agilent Bioanalyzer, pooled, and sequenced on an Illumina NextSeq 500 (paired-end, 38 base-pair reads) to a depth of approximately 30 million reads per sample.
Data alignment
Quality of the sequenced RNA-Seq libraries was assessed using FastQC v0.11.2 (Babraham Bioinformatics). Read quality filtering and adapter trimming were carried out using Trimmomatic v0.39 22. The resulting reads were then aligned and quantified at the transcript level using Salmon23 against GENCODE M20/GRCm38.p6 with the default parameters, and aggregated to the gene level using tximport24.
Differential gene expression analysis and quantification
Gene expression values measured as relative counts were generated using DESeq2 v1.24.0 25, and a minimum threshold of 10 gene counts across all samples was imposed. Differentially-expressed genes were identified in pair-wise comparisons of the four biological conditions (WT_NG; WT_HG; KO_NG; KO_HG) using DESeq2, implementing the apeglm method26 for effect-size estimation. Genes with fold-change ≥ 1.5 and padj ≤ 0.05 were considered significantly down- or up-regulated. Data visualizations were generated using custom R scripts.
Statistical Analysis
All data were represented as mean ± standard error of the mean (SEM). Data were analyzed in GraphPad Prism 9.0 (La Jolla, California) using Student’s t-test, one-way ANOVA, or two-way ANOVA followed by multiple comparison tests, as appropriate. Statistical significance was set at a threshold of p < 0.05.
RESULTS
Islet transcriptional adaptation during prolonged high glucose exposure
To examine the effects of sustained metabolic demand, we used a controlled in vitro model consisting of 6-day islet culture in normal glucose media (11 mM glucose; NG) or high glucose media (25 mM glucose; HG) (Fig. 1a). For a broad overview of high glucose-induced changes to islet gene expression, we performed bulk RNA-Seq analysis of NG- and HG-cultured BclxβWT and BclxβKO islets. High glucose caused major changes to the transcriptional profiles; in BclxβWT islets the expression of 1708 genes (654 down and 1054 up) were altered, and in BclxβKO islets 2182 genes (1013 down and 1169 up) were differentially expressed after HG culture. Of these HG-sensitive transcripts, 725 were altered in both genotypes (Fig. 1b). Focusing first on the common, genotype-independent, response to excess glucose, we performed pathway enrichment analysis, which revealed that the shared transcripts were dominated by changes related to protein and amino acid processing, as well as cellular metabolism (Fig. 1b). Notably, the strongest enrichment was associated with protein processing in the endoplasmic reticulum (ER). Glucose-stimulated insulin biosynthesis increases pressure on the ER protein processing and folding machinery, which can lead to ER stress and apoptosis if the system is overwhelmed. To evaluate islet ER stress status, we compared the shared transcripts to those associated with the gene ontology term “Response to ER stress” (GO:0034976). Of the 725 transcripts, 32 fell in this category and together they indicated a robust up-regulation related to adaptive ER protein homeostasis, including ER-associated degradation (ERAD), protein translation and folding (Fig. 1c). Itpr1, which encodes isoform 1 of the IP3R ER Ca2+ release channel was down-regulated in HG culture. We previously demonstrated that IP3R activity can exacerbate β-cell ER stress and apoptosis29, suggesting that Itpr1 down-regulation is a protective ER response. The pro-apoptotic gene Trib3 was induced in both BclxβWT and BclxβKO islets and HG culture also caused some genotype-specific changes to ER stress genes, but neither genotype showed other defining characteristics of the transition toward ER stress-induced β-cell apoptosis, such as loss of the adaptive unfolded protein response (UPR) or significant induction of Ddit3 (aka Chop)30 (Fig. 1e & Supplemental Fig. 2).
β-Cell glucotoxicity is also mediated by oxidative stress, so we also compared common HG-altered transcripts to the gene ontology term “Oxidative Stress Response” (GO: 0006979). In combination with qPCR, this identified some ROS-associated changes to islet mRNA expression; Hmox1 and Txnip were up-regulated but other major redox-regulated transcripts, including Gpx1, Cat, Sod1, Sod2 and Nfe2l2 were not significantly affected (Fig. 1d,f and not shown). This indicates some activation of oxidative stress, but not sufficiently severe to induce large-scale changes to the expression of antioxidant response genes.
Further suggesting that the high glucose challenge was not overtly toxic, the RNA-Seq analysis did not show a pro-apoptotic shift in Bcl-2 family transcripts31. Major family members were unaltered by HG in BclxβWT islets, whereas BclxβKO islets only showed an increase in anti-apoptotic Bcl2l2 (aka Bclw; padj < 0.01, NG vs HG). Moreover, no significant islet cell death was detected in response to the 6-day high glucose culture, even in the absence of Bcl-xL (Fig. 1g). Using a more toxic challenge with high concentrations of ribose32–34, we established that deletion of Bcl-xL did in fact sensitize β-cells to oxidative stress-related death (Fig. 1h), consistent with its canonical function in restricting Bax- and Bak-mediated apoptosis35.
Overall, this shows that in our model of prolonged glucose excess, the common transcriptional changes in BclxβWT and BclxβKO islets include signs of beginning oxidative stress, but, as a whole, are non-apoptotic and dominated by robust activation of the adaptive UPR.
Effects of high glucose culture on the expression of glycolytic and β-cell identity genes
In addition to ER protein processing, both BclxβWT and BclxβKO islets responded to HG with altered expression of metabolic genes (Fig. 1). The common changes included up-regulation of a number of transcripts in glycolysis (Fig. 2a). Notably, there was also a significant loss of glucose transporter 2 (Slc2a2) and glucokinase (Gck) expression, as well as up-regulation of Ldha and Pdk1 (Figs. 2a,b). Ldha is one of several ‘disallowed’ genes, the repression of which distinguishes mature β-cells and it helps maintain optimal capacity for glucose-stimulated OXPHOS and insulin secretion36. The increase in Ldha expression was not accompanied by significant up-regulation of other core β-cell disallowed genes37 (Supplemental Fig. 3), but together these transcriptional changes indicate a metabolic reprogramming with increased partitioning of glucose-derived pyruvate away from mitochondria toward lactate (Fig. 2c). Of note, the HG-induced upregulation of glycolysis-related genes included Aldh1a3, which is a feature of dedifferentiated and metabolically deficient β-cells38. Closer inspection further showed higher expression of the α-cell-restricted gene Gc39, and reduced levels of key β-cell markers Ins2, Ucn3, Mafa, Nkx6-1, Slc2a2 (Glut2), Glpr1, and Slc30a8, all of which indicates a beginning loss of mature β-cell identity (Figs. 2d,e). Interestingly, Pdx1 levels were higher in BclxβKO islets (padj < 0.001, BclxβKO vs BclxβWT in HG), but overall, the HG-induced transcriptional changes to glycolysis and β-cell identity were similar between the two genotypes. Together these results show that prolonged high glucose exposure causes metabolic reprogramming and beginning loss of the mature β-cell identity, and this happens similarly in BclxβWT and BclxβKO islets.
Effects of excess glucose and Bcl-xL deletion on β-cell Ca2+ signaling and insulin secretion
To determine how Bcl-xL knockout and HG-induced transcriptional changes related to β-cell function, we first compared acute glucose-stimulated Ca2+ signalling in BclxβWT and BclxβKO islets. Voltage-gated Ca2+ entry is a prerequisite signal that triggers and shapes the kinetics of glucose-stimulated insulin secretion3. After 2 days in NG culture, BclxβWT islets responded to 15 mM glucose with a mix of Ca2+ oscillations (47%) and plateau rises (53%), which is expected for a glucose stimulus of intermediate strength. In contrast, all BclxβKO islets responded with plateau steady-state Ca2+ elevations (Fig 3a), which is normally seen at higher, saturating, glucose concentrations. After 2 days of HG culture, islets of both genotypes had modest elevations in baseline Ca2+ and a beginning delay in their return to basal after removal of the stimulus (Figs. 3a). Additionally, HG-cultured BclxβWT islets no longer oscillated and only showed Ca2+ plateaus similar to BclxβKO islets (Fig. 3a). Extending the HG culture to 6 days did not affect the plateau or peak Ca2+ levels in either genotype (Figs. 3b,c). However, the elevation in basal Ca2+ and the delay in recovering to baseline were worsened, and these disruptions to islet Ca2+ kinetics were exacerbated in the absence of Bcl-xL (Figs. 3b,c). Ca2+ on- and off-rates in response to depolarization with KCl were not affected, so the inability to rapidly terminate glucose-induced Ca2+ signaling was not due to impaired extrusion and buffering mechanisms (Fig. 3d).
A transition from oscillatory to plateau Ca2+ responses at intermediate glucose levels may occur if islets are sensitized to the sugar. We therefore tested the responsiveness of NG- and HG-cultured islets by comparing cytosolic Ca2+ during a step-wise glucose ramp, exemplified in Fig. 3e. As quantified in Fig. 3f, HG pre-culture left-shifted the glucose-response profiles of BclxβWT and BclxβKO islets to a similar degree. Importantly, HG-cultured islets also showed amplified insulin secretion across the same range of glucose concentrations (Fig. 3g). Although the two genotypes had similarly sensitized Ca2+ responses, BclxβKO islets secreted significantly more insulin than BclxβWT islets after both NG and HG culture. Islet insulin content was measured after the secretion assay and did not differ by genotype. The contents were decreased in HG cultured islets, but this was accounted for by the amplified insulin release during the glucose ramp (Supplemental Fig. 4).
Collectively, our data thus far show that islets respond to prolonged glucose excess with compensatory sensitization of Ca2+ responses and amplified insulin secretion, despite transcriptional rewiring of glycolysis and the onset of β-cell dedifferentiation. At this stage, islet Ca2+ response kinetics are perturbed, and both Ca2+ dysregulation and insulin hyper-secretion are exacerbated in the absence of β-cell Bcl-xL.
Bcl-xL is needed to preserve mitochondrial transcripts and mitochondrial hyperpolarization in β-cells under high glucose culture
To better understand the specific role of Bcl-xL in the β-cell response to chronic glucose excess, we next focused on the genotype-dependent differences in gene expression. In NG culture, a total of 52 genes were differentially expressed between BclxβWT and BclxβKO islets. High glucose exposure caused a substantial divergence of the transcriptional profiles and the number of differentially expressed genes increased to 234 (Fig. 4a). Only 4 of the differentially expressed transcripts were common to the two culture conditions, including Bcl2l1, which encodes for Bcl-xL and Esr1, which we believe reflects the expression of the CreERTM transgene in BclxβKO islets. Functional annotation and enrichment analysis showed that the genotype-dependent differences following HG culture were dominated by transcripts related to mitochondria (Fig. 4b). In total, 53 genes fell in the top Cell Component GO-term “Mitochondrion” and 50 (94%) of these were down-regulated in BclxβKO islets compared to BclxβWT. Notably, a large number of these genes encode for subunits of electron transport chain (ETC) complexes I through IV, as well as subunits of the ATP Synthase (Complex V) and mitochondrial ribosomal proteins (Fig. 4c). We confirmed the striking culture- and genotype-dependent differences in key OXPHOS transcripts by qPCR (Fig. 4d).
We next compared mitochondrial function by measuring glucose-stimulated OCR and mitochondrial membrane potential (ΔΨm). Pre-culture in high glucose significantly amplified total OCR (OCRtot) in islets from both genotypes across a range of glucose stimuli. Under all conditions, BclxβKO islets consumed significantly more oxygen than BclxβWT islets (Fig. 4e). In agreement with the increased basal OCR, HG pre-culture also increased the basal ΔΨm of dispersed BclxβWT β-cells in 3 mM glucose, but the ΔΨm response to acute stimulation by glucose was not amplified (Fig. 4f). Notably, BclxβKO cells developed a defect in glucose-stimulated ΔΨm hyperpolarization when cultured in HG (Fig. 4f). This suggests Bcl-xL helps β-cells preserve their mitochondrial proton-motive force when challenged by chronic glucose excess. By comparing OCR in the presence and absence of the ATP Synthase inhibitor oligomycin, we quantified the amount of glucose-stimulated oxygen consumption that was coupled to ATP synthesis (OCRATP)18,40. Despite the drastic increase in total OCR, HG culture did not increase OCRATP in either genotype. Consequently, HG notably reduced the fraction of OCRtot that was coupled to ATP production, particularly in BclxβKO islets (Fig. 4e). This indicates that the impaired ΔΨm response in Bcl-xL KO β-cells is a consequence of respiratory uncoupling.
Bcl-xL is required for normal β-cell expression of Tfam and Pgc-1α
Our RNA-Seq analysis (Fig. 4c) showed an intriguing loss of the mitochondrial transcription factor Tfam specifically in HG-cultured BclxβKO islets. Tfam is essential for transcription and replication of the mitochondrial genome, as well as packaging of mitochondrial DNA (mtDNA) into nucleoids41. In β-cells, Tfam is important for mitochondrial function and insulin secretion42,43. We therefore confirmed the BclxβKO-specific loss of Tfam by qPCR (Fig. 5a). Tfam levels are transcriptionally regulated by Nrf-1 and its co-regulator Pgc-1α, which also control the expression of a large number of nuclear genes involved in mitochondrial biogenesis and function, including OXPHOS components. We examined Ppargc1a expression and found that Pgc-1α mRNA levels were increased by HG-culture in BclxβWT islets. In marked contrast, the induction of Pgc-1α transcript by glucose stress was completely absent in BclxβKO islets and β-cells (Figs. 5a). The dysregulation of Pgc-1α and Tfam expression in BclxβKO β-cells suggests Bcl-xL may also be important for the control of mitochondrial mass.
Bcl-xL limits high glucose-induced perturbations of β-cell mitochondrial dynamics and mass
Mitochondrial function, and their adaptation to metabolic demand, are closely interrelated with the regulation of mitochondrial network morphology and total mitochondrial mass. Network connectivity is shaped by the balance of dynamic fusion and fission events44,45, while mass is dictated by the relative amounts of mitochondrial biogenesis versus degradation by macroautophagy and mitophagy46,47. The effects of prolonged glucose excess on β-cell mitochondrial morphofunction and abundance, however, remain unclear. It is also an important unanswered question if Bcl-xL affects mitochondrial dynamics in β-cells, as reported in neurons48.
HG culture significantly increased expression of the mitochondrial fusion regulator Mfn1 exclusively in BclxβKO islets. The mRNA levels of other fusion-mediators Mfn2 and Opa1, as well as the mitochondrial fission regulators Drp1 and Fis1, were not significantly affected by HG challenge in either genotype (Fig. 5b and Supplemental Fig. 5). To determine the overall impact on mitochondrial networking, we used our recent pipeline for confocal analysis of mitochondrial dynamics and the associated Mitochondria Analyzer software tool19,20. For the most accurate comparisons19,21, we generated full 3D reconstructions of the mitochondrial networks and extracted 9 parameters to quantitatively describe the morphology and structural complexity of each individual organelle. In parallel, we performed hierarchical clustering of parameters extracted from independent imaging experiments, which identified 4 primary morphological categories of β-cell mitochondria; designated as Puncta, Tubular, Filamentous, and Highly Complex (see Table 1 in Methods for defining features). We then established a training set and performed an unbiased classification of ∼16,000 individual mitochondria from NG- and HG-cultured BclxβWT and BclxβKO cells into the 4 morphological categories using machine learning algorithms (Fig. 5c and Methods). By calculating the fraction of each cell’s total mitochondrial volume that was mapped to each category, we found that HG culture significantly increased the population of Highly Complex mitochondria in BclxβKO cells (p<0.005, HG vs NG), but not in BclxβWT cells (p=0.36; HG vs NG) (Fig. 5d). Combined with the increase in Mfn1, this indicates an increase in mitochondrial fusion, which was substantiated by a larger average organelle volume, quantified on a per-cell basis (Fig. 5f). Using a parallel coordinates plot we can get an overview of how mitochondrial parameters vary between genotype and culture conditions (Fig. 5e), and this highlights that BclxβKO cells respond to high glucose with the most pronounced changes to all descriptors of mitochondrial morphology and connectivity. Importantly, BclxβKO cells also had a marked increase in total mitochondrial volume that was not seen in BclxβWT cells (Fig. 5f). Because mitochondrial branch diameter and sphericity were reduced (Fig. 5e), it is unlikely that the additional volume reflects stress-induced damage and swelling.
The observed increase in total mitochondrial volume might explain how HG-cultured BclxβKO islets remain glucose-responsive and hyper-secrete insulin, despite a significant loss of mitochondrial transcripts and ΔΨm. Our RNA-Seq analyses did not indicate major differences in TCA cycle components, suggesting the defects might be predominantly ETC-related. If so, an increase in mitochondrial volume should still enhance the formation of mitochondria-derived factors for amplification of insulin secretion49. Indeed, glucose-dependent amplification of KCl-stimulated insulin secretion in the presence of diazoxide was significantly augmented in BclxβKO islets, compared to BclxβWT (Fig. 5g), consistent with a volume-dependent compensation for ETC-specific defects.
Mitochondrial ROS drives transcriptional reprogramming and Bcl-xL-dependent differences in mitochondrial homeostasis
We previously reported that Bcl-2 has functions in β-cell redox control, and that small molecule co-inhibitors of Bcl-2 and Bcl-xL promote β-cell ROS formation11. A specific role of Bcl-xL was not examined, but ROS could conceivably link Bcl-xL to changes in mitochondrial dynamics and function50, as well as mitochondrial transcription and retrograde control of nuclear gene expression47,51–53.
We examined mitochondrial ROS (mitoROS) levels using the superoxide-specific sensor MitoSOX and found that deletion of β-cell Bcl-xL caused a significant mitoROS increase in NG cultured cells, and that HG culture amplified mitoROS in cells of both genotypes (Fig. 6a). MitoROS levels could be normalized using the mitochondria-targeted anti-oxidant MitoTEMPO (Fig. 6a), which substantiates that the mitoROS is mitochondrial and likely ETC-derived. To determine the metabolic consequences of the observed ROS differences, we cultured BclxβWT and BclxβKO islets in NG or in HG with and without the addition of MitoTEMPO and then exposed them to a mitochondrial stress test, which incorporated an acute stimulation with 15 mM glucose (Fig. 6b). As expected, HG pre-culture elevated basal and glucose-stimulated OCR (Fig. 6c). Respiratory control analysis18 substantiated our previous observation that β-cell Bcl-xL deficiency and HG culture impair ATP coupling efficiency, and further showed that this was associated with the development of a major mitochondrial proton leak (Fig. 6d). Consistent with the observed increases in mitochondrial mass (Fig. 5f), Bcl-xL deletion and HG culture also notably amplified spare respiratory capacity. These perturbations to islet respiratory parameters were all mitigated by lowering of mitoROS levels (Fig. 6d). Scavenging mitoROS during HG culture also amplified the already elevated Ca2+ responses to stimulation with low glucose concentrations (Fig. 6e). ROS has roles in physiological signal transduction, but this result indicates that mitoROS limits, rather than mediates, sensitization of the triggering Ca2+ signal for glucose-stimulated insulin secretion in HG-cultured islets.
We next assessed the role of mitoROS in the HG-induced changes to β-cell gene expression. Remarkably, all major glucose-induced transcriptional changes we assessed by qPCR were partially or fully reversed by normalization of mitochondrial ROS (Fig. 7a,b and Supplemental Fig. 6). This included rescue of the HG-induced changes related to glycolytic flux and β-cell identity (Gck, Ldha, Pdk1, Ins2, Mafa) that were common to both genotypes, as well as the loss of Tfam, and ETC components (Ndufb8, Sdhb, Cox4l1, Atp5e) that was specific to BclxβKO islets.
DISCUSSION
In this study, we characterized the pancreatic β-cell response to sub-lethal levels of glucose stress, and used conditional gene deletion to determine the role of the anti-apoptotic protein Bcl-xL in these processes. In both wild-type and Bcl-xL knockout islets, prolonged glucose excess caused a robust transcriptional reprogramming of metabolism toward increased glycolysis and beginning loss of the mature β-cell identity. These changes to gene expression in large part resemble the signatures of hyperglycemia-induced β-cell dedifferentiation in partially pancreatectomized rats54,55. Interestingly, our results showed that in glucose-stressed cells, the transcriptional loss of β-cell identity occurs while there is still compensatory β-cell hyper-function, and precedes the development of maladaptive ER stress. This indicates that loss of the mature differentiated β-cell phenotype manifests remarkably early in the progression of glucotoxicity. Whether this sequence of events differs under other forms of diabetogenic stress is not clear.
Maladaptive ER stress, has been proposed as a primary event preceding oxidative and inflammatory stress in T2D30. By all indications, the glucose-induced changes to β-cell UPR progressed similarly in BclxβWT and BclxβKO islets, suggesting that Bcl-xL does not modulate the susceptibility of β-cells to ER stress, unlike the pro-apoptotic Bax and Bak56,57. Rather, we identified a mitochondria-specific defect in the absence of Bcl-xL. Prolonged high-glucose exposure amplified respiration, but Bcl-xL-deficient β-cells developed mitochondrial dysfunction with reduced coupling of ETC flux to ΔΨm hyperpolarization and ATP synthesis. Importantly, our results also revealed that Bcl-xL is essential for normal β-cell transcription of OXPHOS components and mitochondrial ribosomes under chronic glucose excess. Overall, the loss of mitochondrial transcripts in glucose-stressed BclxβKO islets is similar to the metabolic gene expression signatures seen 2 weeks after the onset of hyperglycemia in non-obese diabetic mice6, and during late stages of diabetes progression in the non-obese GK rat8. BclxβKO islets also showed an exacerbated perturbation of basal and glucose-stimulated Ca2+ response kinetics. Together, this suggests that as glucose levels increase, Bcl-xL counteracts mitochondrial decompensation, and thereby likely the development of a detrimental feed-forward loop of β-cell dysfunction and worsening hyperglycemia.
It is noteworthy that BclxβKO islets remained hyper-responsive despite marked loss of mitochondria-related gene expression and beginning uncoupling of respiration from ΔΨm. The ability of BclxβKO islets to uphold Ca2+ and secretory responses in the face of beginning mitochondrial dysregulation was associated with profound changes to mitochondrial morphology and volume. Previous work has shown that mitochondrial fusion serves as a protective response to sustain survival under nutrient stress58. Our results indicate that mitochondrial fusion and expansion of total mitochondrial mass also are integral to functional compensation of β-cells, prior to the failure and mitochondrial fragmentation that characterizes β-cells in rodents and humans with established T2D4–6. In HG-cultured Bcl-xL KO cells, the total mitochondrial volume increased despite the absence of glucose-induced Ppargc1a expression, which may reflect the presence of Pgc-1α-independent mechanisms for maintenance of mitochondrial mass in β-cells59. Alternatively, mitochondrial hyper-fusion in Bcl-xL KO β-cells might contribute to the expansion of total mass by counteracting mitochondrial degradation by autophagy and mitophagy60.
Using a mitochondria-targeted antioxidant, we identified sub-lethal formation of mitoROS as a primary driver of transcriptional changes to both glycolysis and β-cell identity in glucose-stressed islets. In both NG and HG culture, Bcl-xL-deficient β-cells had higher mitoROS levels than wild-type. Importantly, mitochondrial antioxidants normalized respiratory coupling and restored nuclear-encoded mitochondrial gene expression in HG-cultured BclxβKO islets. This indicates that Bcl-xL, protects mitochondrial homeostasis in glucose-stressed β-cells by limiting the levels of mitoROS. We previously reported that Bcl-2 dampens the physiological formation of peroxides in β-cells11, but a similar role for Bcl-xL was not examined. The excess mitoROS in Bcl-xL KO β-cells is likely produced in the ETC as a consequence of the overall increase in mitochondrial respiration. It remains to be established how deletion of Bcl-xL accelerates β-cell metabolic activity and insulin secretion. The transcriptional profiles of BclxβKO and BclxβKO islets in normal culture did not indicate differences in major metabolic pathways, so changes to protein-protein interactions likely play a role. In neurons, Bcl-xL increases metabolic efficiency by limiting an ion leak conductance through the F1Fo ATPase complex61. Loss of such a stabilizing function may well contribute to proton leak and respiratory uncoupling in Bcl-xL KO β-cells, but it would be expected to impair glucose-stimulated Ca2+ entry, insulin secretion and mitoROS production, in contrast to what we observe. Additional mechanisms are therefore likely at play, and future metabolomics profiling may help clarify the functions of Bcl-xL in β-cell energetics.
In summary, we have identified important functions for Bcl-xL in preserving mitochondrial integrity in pancreatic β-cells under non-apoptotic levels of metabolic stress. Our findings provide new insights into non-apoptotic roles Bcl-2 family survival proteins in the control of organelle physiology. We further revealed central role for mitoROS in both the functional and transcriptional impact of chronic glucose excess on β-cells. Future work is warranted to clarify the relationship between Bcl-xL, mitoROS, and mitochondrial homeostasis in the pathogenesis and treatment of T2D.
DUALITY OF INTEREST
The authors report no conflicts of interest.
AUTHOR CONTRIBUTIONS
D.J.P. and R.S. contributed equally to this work. D.J.P., R.S. and D.S.L. designed the study and wrote the paper. D.J.P., R.S., B.V., A.Z.L.S., Y.Z., A.C. and D.S.L. performed experiments and analyzed data. D.J.P., R.S., B.V., A.Z.L.S., Y.Z., A.C., B.G.H. and D.S.L interpreted data and edited the paper.
ACKNOWLEDGEMENTS
This work was supported by an Operating Grant to D.S.L. from the Canadian Institutes for Health Research (CIHR; MOP-119537). D.S.L. was supported by JDRF (CDA 2-2013-50) and BC Children’s Hospital Research Institute (BCCHR). R.S. was supported by a Canada Graduate Scholarship from CIHR. We acknowledge Dr. Jingsong Wang (BCCHR Imaging Core), as well as Mei Tang and Mitsuhiro Komba, (BCCHR) for expert technical assistance.
Footnotes
Abbreviations: ΔΨm, mitochondrial membrane potential; DZ, diazoxide; ER, Endoplasmic Reticulum; ERAD, ER-associated degradation; ETC, Electron Transport Chain; FCCP, carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone; GO, Gene Ontology; HG, high glucose; KRB, Kreb’s Ringer’s Buffer; mitoROS, mitochondrial ROS; MT, MitoTempo; mtDNA, mitochondrial DNA; MTG, MitoTracker Green; NG, normal glucose; OCR, oxygen consumption rate; OXPHOS, oxidative phosphorylation; T2D, type 2 diabetes; TM, tamoxifen; TMRE, tetramethylrhodamine, ethyl ester; UPR, unfolded protein response.