Biochemical Journal

Research article

Metabolomic analyses reveal profound differences in glycolytic and tricarboxylic acid cycle metabolism in glucose-responsive and -unresponsive clonal β-cell lines

Peter Spégel, Siri Malmgren, Vladimir V. Sharoyko, Philip Newsholme, Thomas Koeck, Hindrik Mulder

Abstract

Insulin secretion from pancreatic β-cells is controlled by complex metabolic and energetic changes provoked by exposure to metabolic fuels. Perturbations in these processes lead to impaired insulin secretion, the ultimate cause of T2D (Type 2 diabetes). To increase our understanding of stimulus–secretion coupling and metabolic processes potentially involved in the pathogenesis of T2D, a comprehensive investigation of the metabolic response in the glucose-responsive INS-1 832/13 and glucose-unresponsive INS-1 832/2 β-cell lines was performed. For this metabolomics analysis, we used GC/MS (gas chromatography/mass spectrometry) combined with multivariate statistics. We found that perturbed secretion in the 832/2 line was characterized by disturbed coupling of glycolytic and TCA (tricarboxylic acid)-cycle metabolism. The importance of this metabolic coupling was reinforced by our observation that insulin secretion partially could be reinstated by stimulation of the cells with mitochondrial fuels which bypass glycolytic metabolism. Furthermore, metabolic and functional profiling of additional β-cell lines (INS-1, INS-1 832/1) confirmed the important role of coupled glycolytic and TCA-cycle metabolism in stimulus–secretion coupling. Dependence of the unresponsive clones on glycolytic metabolism was paralleled by increased stabilization of HIF-1α (hypoxia-inducible factor 1α). The relevance of a similar perturbation for human T2D was suggested by increased expression of HIF-1α target genes in islets from T2D patients.

  • hypoxia-inducible factor (HIF)
  • mitochondria
  • pancreatic islet
  • insulin
  • Type 2 diabetes (T2D)

INTRODUCTION

Failure of the pancreatic β-cells to release insulin appropriately is a major pathogenetic abnormality in T2D (Type 2 diabetes). Insulin secretion from β-cells is controlled by two different mechanisms: the triggering (KATP-dependent) and the amplifying (KATP-independent) pathways [1]. While the triggering pathway has been thoroughly characterized [2,3], the mechanisms and signals underlying the amplifying pathway remain largely unknown [4,5]. Yet several metabolites, mainly originating from mitochondrial metabolism, have been implicated as potential coupling factors in the amplifying pathway of GSIS (glucose-stimulated insulin secretion). These include long-chain acyl-CoAs [6], glutamate [7] and NADPH [8,9].

In the present study, the metabolome of two clonal β-cell lines, the glucose-responsive 832/13 and the glucose-unresponsive 832/2 lines, both of which were derived from the INS-1 rat insulinoma cell line [10], were examined. This enabled us to investigate possible differences in metabolic regulation that underlie perturbed function of the 832/2 cells. Expression of several glycolytic and TCA (tricarboxylic acid)-cycle enzymes has previously been found to differ between these cell lines [11]. Most strikingly, the 832/2 cells were found to express LDHA (lactate dehydrogenase A) and release lactate. Hence, the glucose-unresponsive 832/2 line relied heavily on glycolytic metabolism, while the robust insulin secretor, the 832/13 line, exhibited strong glucose concentration-dependent coupling of cytosolic and mitochondrial fuel metabolism. Moreover, it has previously been shown that 832/13 cells are characterized by an active anaplerotic pathway, which is less pronounced in 832/2 cells [12]. Nevertheless, a detailed analysis of the impact of these metabolic differences on the levels of individual metabolites and how they are controlled has not yet been performed.

To date, the bulk of investigations on the coupling of β-cell metabolism to insulin secretion has been performed in a univariate fashion. One or a few metabolites or metabolic enzymes have been studied at a time. Owing to the complexity of the metabolic networks, univariate investigations are probably insufficient. An unbiased characterization of metabolism is therefore warranted. Thus in the present study, GC/MS (gas chromatography/mass spectrometry)-based metabolomics [1315] combined with multivariate statistical analyses [16] were applied to investigate the metabolome. These metabolomic analyses generated hypotheses, which we aimed to verify with functional studies and analysis of human islets.

A prominent finding was the lack of response in levels of late glycolytic and TCA-cycle intermediates in the unresponsive 832/2 cell line, despite an increase in extracellular glucose. The perturbed coupling of cytosolic and mitochondrial metabolism could partially be circumvented by specifically stimulating mitochondrial metabolism. This manoeuvre reinstated fuel-stimulated mitochondrial respiration and insulin secretion. Interestingly, we found that HIF-1α (hypoxia-inducible factor 1α), a transcription factor regulating the balance between glycolytic and mitochondrial metabolism, was stabilized in the glucose-unresponsive cells.

EXPERIMENTAL

Metabolomics

The clonal β-cell lines INS-1 832/2 and INS-1 832/13 were stimulated with either 2.8 or 16.7 mM glucose for 1 h followed by an unbiased analysis of the metabolome using GC/MS. The generated data were normalized and evaluated using multivariate statistics to highlight alterations in metabolite levels unique to, or shared between, the two cell lines. To validate the finding in the 832/2 and 832/13 clones, a metabolite profiling of the glucose-responsive INS-1 and the glucose-unresponsive INS-1 832/1 clones was undertaken. Details of the metabolomics protocols are found in the Supplementary information (at http://www.BiochemJ.org/bj/435/bj4350277add.htm).

Insulin secretion

The INS-1, 832/1, 832/2 and 832/13 clones were cultured in 24-well culture plates until 90–95% confluency was reached, whereafter insulin was assayed as previously described in detail [10]. Insulin secretion was measured after incubation for 1 h in 2.8 mM glucose, 16.7 mM glucose or 10 mM leucine and 10 mM glutamine.

Oxygen consumption

The OCR (oxygen consumption rate) was measured in the 832/2 and 832/13 clones using the XF (extracellular flux) analyser XF24 (Seahorse Bioscience), as previously described in detail [11]. Following a pre-incubation at 2.8 mM glucose for 2 h, the OCR was assayed at 2.8 mM glucose subsequent to a transition to 16.7 mM glucose or 2.8 mM glucose together with 10 mM leucine and 10 mM glutamine. OCRs were first normalized by protein content and then by the average basal OCR at 2.8 mM glucose. An area under the curve analysis was performed for fuel-stimulation conditions.

Glucose uptake

The cells were prepared as for analysis of insulin secretion [10]. After removing the HBSS (Hepes-balanced salt solution; 114 mM NaCl, 4.7 mM KCI, 1.2 mM KH2PO4, 1.16 mM MgSO4 20 mM Hepes, 2.5 mM CaCl2, 25.5 mM NaHCO3 and 0.2% BSA, pH 7.2), 500 μl of HBSS containing 3-O-methyl-D-[3H(N)]glucose (specific activity, 80.2 Ci/mmol; PerkinElmer Life Sciences) and 3-O-methyl-D-glucose at final concentrations of 2.8 and 16.7 mM respectively, were added to each well. Cells were incubated for 10 min at 37 °C. Next, 250 μl of lysis buffer containing 200 mM NaCl, 2 mM EDTA, 50 mM Tris/HCl (pH 7.4) and 1% (w/v) SDS were added. Subsequently, 150 μl of lysate were transferred into scintillation vials. Protein content was determined by the BCA (bicinchoninic acid) method. 3-O-Methyl-D-[3H(N)]glucose content was measured by liquid scintillation spectrometry [17].

Western blotting

Whole cells were solubilized in homogenization buffer [100 mM Hepes, 9 M urea, 1% (v/v) Triton X-100 and 2 mM EDTA, pH 7.2] mixed 1:100 (v/v) with protease inhibitor cocktail (Sigma). Protein content was determined by the BCA method. Samples were mixed 1:10 (v/v) with loading buffer [100 mM Hepes, 10% (w/v) SDS, 10% (v/v) dithiothreitol and 20% (v/v) glycerol, pH 7.2]. A 40 μg amount of protein was run on an SDS/PAGE (8% gel), and subsequently blotted on to PVDF membranes. HIF-1α protein was detected with a primary mouse anti-HIF-1α monoclonal antibody (Abcam) in dilution 1:500 (v/v). β-Tubulin was detected with a primary rabbit polyclonal antibody (Santa Cruz Biotechnology) in dilution of 1:400 (v/v). Horseradish peroxidase-coupled goat-anti-rabbit IgG (1:8000, v/v), goat-anti-mouse IgG (1:6000, v/v), and donkey-anti-goat IgG (1:8000, v/v) (Santa Cruz Biotechnology) were used as secondary antibodies. Blots were developed with ECL (enhanced chemiluminescence) and detection was by Amersham Hyperfilm ECL (Amersham Biosciences). Films were scanned using the Bio-Rad GS-800 Calibrated Densitometer (Bio-Rad). Measurements obtained for HIF-1α were corrected for protein loading with β-tubulin. For loading control, blots were stripped with 2% (v/v) SDS, 100 mM 2-mercaptoethanol and 62.5 mM Tris/HCl, pH 6.7.

Human islets

Human pancreatic islets from 55 non-diabetic and nine T2D deceased donors were obtained from the Human Tissue Laboratory at Lund University Diabetes Centre. The 29 male and 26 female non-diabetic donors were aged 56.7±9.8 years, had a BMI (body mass index) of 25.9±3.6 kg/m2 and HbA1c (glycated haemoglobin) of 5.7±0.8%. The five male and four female T2D donors were aged 57.0±13.1 years, had a BMI of 28.5±4.7 kg/m2 and HbA1c of 7.3±1.2%. Islets were prepared by collagenase digestion and density gradient purification. After isolation, islets were cultured free floating in CMRL 1066 culture medium (ICN Biomedicals) supplemented with 10 mmol/l Hepes, 2 mmol/l L-glutamine, 50 μg/ml gentamicin, 0.25 μg/ml Fungizone (Gibco BRL), 20 μg/ml ciprofloxacin (Bayer Healthcare) and 10 mmol/l nicotinamide at 37 °C (5% CO2) prior to RNA and DNA preparation. The donor, or her/his relatives, upon admission to the ICU (intensive care unit) had given their consent to donate organs and the local ethics committees approved the protocols. The procedure adhered to the Declaration of Helsinki (2000) and the World Medical Association.

Total RNA was isolated with the AllPrep DNA/RNA Mini Kit (Qiagen). The microarrays were performed following the Affymetrix standard protocol. Briefly, 200 ng of total RNA were processed following the GeneChip® Expression 3′-Amplification Reagents One-cycle cDNA synthesis kit instructions (Affymetrix) to produce double-stranded cDNA. This was used as a template to generate biotin-targeted cRNA following manufacturer's specifications. A 15 μg aliquot of the biotin-labelled cRNA was fragmented to strands between 35 and 200 bp in length, 10 μg of which was hybridized on to the GeneChip® Human Gene 1.0 ST whole transcript based assay overnight in the GeneChip® Hybridization oven 6400 using standard procedures. The arrays were washed and stained in a GeneChip® Fluidics Station 450. Scanning was carried out with the GeneChip® Scanner 3000 and image analysis was performed using GeneChip® Operating Software. The array data were summarized and normalized with RMA (robust multi-array analysis) method using the Expression Console software (Affymetrix).

Statistical analysis

All data are shown as means±S.D. or S.E.M. for the indicated number of experiments. A Mann–Whitney U test was used to compare data from human islets. Unless stated otherwise, Student's t test was used when comparing two groups. The Kruskal–Wallis test in combination with Mann–Whitney followed by Bonferroni's test post hoc was applied for the analysis of HIF-1α protein expression.

RESULTS

Metabolomics

Multivariate statistical calculations were performed to identify which metabolites out of the 164 detected were uniquely regulated by the glucose stimulation in the INS-1 832/2 and the INS-1 832/13 clones (see Supplementary information). These analyses revealed a unique regulation of the late glycolytic and TCA-cycle intermediates in the 832/13 clone (Figures 1A and 1B). Whereas glucose-6-phosphate was found to exhibit a glucose-stimulated increase in both cell lines, downstream intermediates 3-phosphoglycerate, 2-phosphoglycerate, phosphoenolpyruvate and pyruvate were found to increase exclusively in the glucose-responsive 832/13 clone. This lack of response in the 832/2 clone was reflected also by the levels of the TCA-cycle intermediates; levels of all measured TCA-cycle intermediates increased in the 832/13 clone whereas they were unaltered in the 832/2 clone (Figure 1B). The fold of the glucose-stimulated increase in lactate was similar in the two clones (Figure 1A), although the levels were significantly higher in the unresponsive clone (see Supplementary Figures S4 and S8 at http://www.BiochemJ.org/bj/435/bj4350277add.htm). The level of the pentose phosphate shunt intermediate ribose-5-phosphate rose more in the glucose-responsive 832/13 clone.

Figure 1 Glucose-stimulated fold changes in metabolite levels

Fold changes of metabolite levels after glucose stimulation were calculated by dividing the peak areas for the metabolite levels at 16.7 and 2.8 mM glucose derived from the glucoseunresponsive INS-1 832/2 and the glucose-responsive INS-1 832/13 cell lines respectively. (A) Glycolytic and related metabolites. (B) TCA-cycle and related intermediates. The data are represented as means±S.E.M. Statistical comparisons were made by Student's t test (*P<0.05, **P<0.01, ***P<0.001). Rib5P, ribose-5-phosphate; Glu6P1, glucose-6-phosphate peak 1; Glu6P2, glucose-6-phosphate peak 2; GlyA3P, 3-phosphoglycerate; GlyA2P, 2-phosphoglycerate; PEP, phosphoenol pyruvate; Pyr, pyruvate; Lac, lactate; Cit, citrate; Aco, aconitate; AKGA, α-ketoglutarate; Succ, succinate; Fum, fumarate; Mal, malate.

Oxygen consumption

Given the lack of rise in TCA-cycle intermediates in the 832/2 clone in response to glucose, we were interested to find out whether this had any further functional implications. To this end we chose to assess oxygen consumption, which reflects overall mitochondrial activity. Based on an area under the curve analysis, glucose stimulation caused a sustained 1.84±0.26-fold (n=7) increase in OCR in the 832/13 clone, whereas no change in OCR was observed upon glucose stimulation of the 832/2 clone (Figures 2A and 2C). However, leucine and glutamine increased the OCR in both clones; OCR was significantly increased 1.81±0.18-fold (n=6) and 1.26±0.07-fold (n=6) in 832/13 and 832/2 cells respectively (Figures 2B and 2C), compared with respiration at 2.8 mM glucose alone.

Figure 2 Respiration and insulin secretion in glucose-responsive 832/13 and glucose-unresponsive 832/2 cells

(A and B) OCR in INS-1 832/13 (closed triangles) and 832/2 (open triangles) cells after addition of metabolic fuels: (A) 16.7 mM glucose or (B) 10 mM leucine+10 mM glutamine. (C) Fuel-stimulated OCR during the stimulation period normalized to protein content and basal OCR at 2.8 mM glucose; calculated based on an area under curve (AUC) analysis [28]. (D) Fold change of insulin secretion in INS-1 832/13 and INS-1 832/2 β-cells at 16.7 mM glucose and 10 mM leucine+10 mM glutamine relative 2.8 mM glucose. Data are presented as means ± S.E.M. (*P<0.05, **P<0.01, ***P<0.001).

Insulin secretion

The lack of a glucose-stimulated response in levels of TCA-cycle intermediates in the glucose-unresponsive 832/2 clone probably constrains insulin secretion via both the triggering pathway, which is dependent on the mitochondrial ATP production, and the amplifying pathway, which has been suggested to be coupled to TCA-cycle metabolism. Basal insulin secretion at 2.8 mM glucose was 15.9±1.8 (n=8) and 8.39±1.96 (n=6) ng/mg per h for 832/13 and 832/2 cells respectively. Indeed, while stimulation with 16.7 mM glucose yielded a 12.6±1.3-fold (n=8) increase of insulin secretion in the glucose-responsive 832/13 cells, no rise in GSIS was observed in the glucose-unresponsive 832/2 cell line (Figure 2D). Leucine and glutamine, on the other hand, significantly stimulated insulin secretion in 832/2 cells [2.88±0.22-fold (n=6) increase compared with the basal level], whereas the same stimulus yielded a 14.1±5.5-fold (n=6) increase in the glucose-responsive 832/13 cell line (Figure 2D).

Metabolite profiling of additional clonal β-cell lines

The results generated from the study of the glucose-responsive 832/13 clone and the glucose-unresponsive 832/2 clone suggested that insufficient coupling of glycolytic and TCA-cycle metabolism may be a factor underlying perturbed GSIS. To investigate whether this phenotype is unique for the glucose unresponsiveness of the 832/2 clone, or a more general phenomenon in β-cells, two additional β-cell lines were investigated. To this end, the glucose-responsive INS-1 line and the glucose-unresponsive INS-1 832/1 subclone were examined. A targeted metabolite profiling revealed a significantly lower fold change in glucose-stimulated alterations in glycolytic and TCA-cycle metabolite levels in the unresponsive clone (Figure 3), reminiscent of the metabolic phenotype of the 832/2 clone (Figures 1A and 1B).

Figure 3 Targeted metabolite profiling of the glucose-responsive INS-1 parental clone and the glucose-unresponsive INS-1 832/1 subclone

A targeted metabolite profiling was performed to investigate whether the observed metabolic perturbation was present also in other clonal β-cell lines. Fold changes were calculated and compared as described in the legend to Figure 1; (*P<0.05, **P<0.01, ***P<0.001). Rib5P, ribose-5-phosphate; Glu6P1, glucose-6-phosphate peak 1; Glu6P2, glucose-6-phosphate peak 2; GlyA3P, 3-phosphoglycerate; GlyA2P, 2-phosphoglycerate; PEP, phosphoenol pyruvate; Lac, lactate; Cit, citrate; Aco, aconitate; AKGA, α-ketoglutarate; Succ, succinate; Fum, fumarate; Mal, malate.

Functional characterization of additional clonal β-cell lines

Next, the functional consequences of the reduced response in the glycolytic and TCA-cycle intermediates upon glucose stimulation were assessed. Glucose was found to stimulate insulin secretion 6.6±2.6-fold (n=3) in the INS-1 parental clone, whereas insulin secretion from the 832/1 subclone was not stimulated by the hexose (1.2±0.3-fold, n=3). To investigate whether stimulus–secretion coupling was intact and the observed lack of response in the 832/1 clone was due to insufficient coupling of glycolytic and TCA-cycle metabolism, insulin secretion in response to leucine and glutamine was also assessed. Indeed, leucine and glutamine provoked a significant (P<0.05) 7.8±2.4-fold and a 3.1±1.0-fold increase in insulin secretion in the INS-1 and the 832/1 clone respectively.

Glucose uptake

To investigate whether differences in glucose uptake could account for the observed metabolic differences between the glucose-responsive and -unresponsive β-cell clones, uptake of the non-metabolizable analogue of D-glucose, 3-O-methyl-D-glucose, was assessed. No significant differences between the four clones could be observed (see Supplementary Figure S9 at http://www.BiochemJ.org/bj/435/bj4350277add.htm), indicating that differences in glucose uptake alone cannot explain the perturbed metabolic response in the glucose-unresponsive clones.

Stabilization of HIF-1α

The metabolomics data suggested that the lacking response of the 832/2 clone to glucose stimulation may be caused by impaired aerobic glycolysis concurrent with lactate production and a perturbation in the coupling between cytosolic and TCA-cycle metabolism. This, taken together with the fact that the cell lines are tumour-derived [10], suggested involvement of HIF-1α. Stabilization of this protein may account for a switch of cellular metabolism to dependence on glycolysis, the so called Warburg effect [18]. To explore this possibility, we determined the presence of HIF-1α by Western blotting. Indeed, while HIF-1α was present in all INS-1 clones, its protein levels were, for example, 227±30% (P= 0.006) higher in the glucose-unresponsive 832/2 clone, in comparison with the glucose-responsive 832/13 clone (Figures 4A and 4B).

Figure 4 Stabilization and functional relationships of HIF-1α with metabolic and functional parameters in clonal β-cell lines

HIF-1α stabilization was examined by Western blotting as described in the Experimental section. n=3–6; data were compared by Kruskal–Wallis test in combination with Mann–Whitney test followed by Bonferroni's test post hoc. (A) Western blot data. (B) Quantified levels of HIF-1α. (C) Stabilization of HIF-1α in relation to the ratio of α-ketoglutarate (AKGA) to succinate (Succ) (calculated as the ratio of their fold changes to stimulatory glucose). (D) Stabilization of HIF-1α in relation to fold changes of citrate (Cit) to pyruvate (Pyr) (see C). (E) Relationship between the fold change of GSIS and HIF-1α stabilization. Data are presented as means ± S.E.M. (**P<0.01).

It has previously been shown that a low level of α-ketoglutarate in the presence of a high level of succinate stabilizes HIF-1α [19,20]. To assess whether such a mechanism is operating in the clonal cell lines, we plotted the HIF-1α protein levels against the ratios of α-ketoglutarate and succinate (fold responses of the metabolite levels to glucose stimulation were employed; Figure 4C). Indeed, we observed that the lower the ratios of α-ketoglutarate to succinate, the higher the protein level of HIF-1α. Next, we examined whether the change in HIF-1α protein levels coincided with metabolic changes. To this end, we plotted HIF-1α protein levels against the ratios of citrate to pyruvate (Figure 4D). The rationale for this analysis is that this ratio may reflect coupling of glycolysis and TCA-cycle activities. A low ratio would imply a low level of TCA-cycle activity, suggesting that coupling of glycolysis and TCA-cycle is perturbed. The plot clearly shows that high HIF-1α protein levels coincided with a low ratio of citrate to pyruvate in the clonal β-cell lines. This translated into impaired GSIS, as shown in Figure 4(E).

Expression of HIF-1α target genes in human islets

Stabilization of HIF-1α can establish a pseudohypoxic metabolic phenotype under normoxic conditions. This involves primarily increased expression of glycolytic enzymes. Thus to investigate whether our findings in the clonal cell lines are of relevance for the pathogenesis of T2D in humans, we first analysed expression of HIF-1α in human islets. Although HIF-1α was expressed at the mRNA level, we did not observe any differences between human islets from healthy controls and patients with T2D (results not shown). This is in agreement with the fact that signalling through HIF-1α requires stabilization on the protein level, a phenomenon which occurs independent of changes in mRNA level of HIF-1α [21]. Instead, to ascertain whether regulation by HIF-1α occurs in human islets, we examined expression of HIF-1α target genes in human islets from healthy controls and patients with T2D. Indeed, we found increased mRNA levels for hexokinase 2, LDHA, and phosphofructokinase-2/fructose-2,6-bisphosphatase-3. Unexpectedly, there were minor, but significant, decreases in pyruvate dehydrogenase kinase-1 (~13%) and enolase (~22%) mRNA levels (Figure 5).

Figure 5 Islet mRNA expression from patients with T2D and healthy controls

Expression of islet mRNA, derived from microarray analysis, of hexokinase 2 (HK2), enolase (ENO2), 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase-3 (PFKFB3), pyruvate dehydrogenase kinase-1 (PDK1) and LDHA. Data are given as means ± S.D. and were compared with a Mann–Whitney U test; (*P<0.05, **P<0.01, ***P<0.001).

DISCUSSION

The complexity of β-cell metabolism prompts novel analytical approaches that allow analyses of more than one or a few metabolites at time. In the present study, metabolomics allowed the simultaneous analysis of 164 putative metabolites of which 44 could be identified. Analyses of four clonal β-cell lines with variable glucose responsiveness pinpointed a fundamental disruption of metabolism in the glucose-unresponsive clones. The disruption appeared to be at the level of coupling glycolysis to TCA-cycle metabolism. As metabolism moved down the path of glycolysis, the weaker the glucose-stimulated response in metabolite levels became. In fact, even lactate, a metabolite not thought to be compatible with robust β-cell function, was increasingly produced in glucose-unresponsive cell lines [11].

To explore whether this observed metabolic perturbation impacted mitochondrial activity, we analysed OCR in the presence of either glucose, which requires glycolytic metabolism, or a combination of leucine and glutamine, which directly activates TCA-cycle metabolism. In agreement with the metabolomics data, glucose was ineffective to stimulate oxygen consumption in glucose-unresponsive cells with the observed block of coupling between glycolysis and TCA-cycle metabolism. In contrast, fuels that were directly metabolized in the TCA-cycle also increased oxygen consumption in glucose-unresponsive cells, albeit not to the same extent as in glucose-responsive cells. This suggests that TCA–cycle metabolism and oxidative phosphorylation in the glucose-unresponsive cell lines have adapted to a metabolic situation where the provision of fuels from glycolysis is constrained. In support of this notion, we have previously shown that respiratory complex activities, and protein expression of key subunits of the complexes, are decreased in 832/2 compared with 832/13 cells [11]. Nevertheless, when mitochondrial fuels are provided, a metabolic response from mitochondria can still be provoked. Accordingly, fuels metabolized directly in the TCA-cycle were able to stimulate insulin secretion also in the clones totally lacking a secretory response to glucose.

At this point, the question remained how the adaptive transition from coupled glycolytic and TCA-cycle metabolism to an increasing dependence on glycolysis was regulated. As we previously observed [11], metabolism in the 832/2 line exhibits Warburg-like features. Moreover, the metabolite profiling revealed that TCA-cycle intermediates, particularly succinate, that can inhibit proline hydroxylases when coinciding with low levels of α-ketoglutarate, were increasingly abundant in cells with low glucose responsiveness. Proline hydroxylases hydroxylate HIF-1α, priming it for proteolytic degradation. The inhibition of this process is an important component of the Warburg transition establishing a pseudohypoxic normoxic condition. HIF-1α serves as a transcription factor, enhancing the expression of genes in the glycolytic pathway. Hereby energy homoeostasis is maintained under hypoxic conditions, a situation commonly found in tumour-derived cells even under normoxic conditions. Indeed, the cell lines examined in the present study are of tumour origin, originally derived from a transplantable rat insulinoma [22]. Accordingly, we found that HIF-1α protein was present in all of the clonal cell lines, being most abundant in the cell lines with the lowest level of glucose responsiveness. In support of this observation we found that increased levels of HIF-1α protein coincided with a perturbed coupling of glycolysis and TCA-cycle metabolism, as evident from the citrate to pyruvate ratio. Moreover, higher HIF-1α protein levels corresponded to impaired GSIS. Collectively, these observations suggest that HIF-1α is a key regulator of the balance between glycolytic and TCA-cycle metabolism in β-cells. In this capacity, HIF-1α could serve as a regulator of β-cell stimulus–secretion coupling. However, while our experiments have established a relationship between HIF-1α and glucose responsiveness, it remains to be shown that this is a causal effect. Supporting a causal relationship are recent findings showing that genetic targeting of the von Hippel–Lindau factor in mouse β-cells stabilizes HIF-1α, resulting in impaired glucose tolerance and perturbed insulin secretion [23]. Yet another study shows that HIF-1α present at low levels in mouse β-cells under normoxic conditions is required for proper β-cell function [24]. This circumstance could explain why we also found HIF-1α protein in the glucose-responsive cell lines, albeit at very low levels.

The relevance of our findings for the pathogenesis of T2D is supported by studies of animal models for the disease. For instance, lactate production has been shown to be increased in the GK (Goto–Kakizaki) rat [25]. Moreover, it has previously been shown that leucine-stimulated glutamate metabolism is intact in the GK rat and that the combination of these amino acids stimulates insulin secretion [26]; the latter is reminiscent of what we observed in the glucose-unresponsive cell lines. In fact, although patients with T2D exhibit severely perturbed GSIS, glutamine/leucine-stimulated insulin secretion is largely preserved [27]. Substantiating our findings on HIF-1α, it has previously been reported that the protein is expressed in human islets [24]. Using our repository of human islets, we could confirm this finding. However, there was no difference between patients with T2D and healthy controls. This was of no surprise given that HIF-1α mRNA levels are not thought to contribute to its regulatory function. Instead, this is controlled by its stabilization at the protein level. Therefore we explored expression of HIF-1α target genes in human islets. Indeed, we found that some of the known targets of this transcription factor, e.g. hexokinase 2 LDHA, and 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase-3 were up-regulated in islets from patients with T2D. Although these data are compelling, there should be call for caution. The human islets used in the present study, as well as in any other studies, have been cultured for different lengths of time. In this situation, the circulation in the islets has been compromised. This implies that hypoxia may prevail in these islets, particularly in the core of the islets. Clearly, such hypoxia could stabilize HIF-1α. This notwithstanding, we did observe a difference between healthy controls and patients with T2D.

In conclusion, we have used metabolomics analysis to identify a profound perturbation in stimulus–secretion coupling in clonal β-cells. The metabolite profile suggested to us that HIF-1α could play a role in the observed transition of metabolism to dependence on glycolysis (Figure 6). The functional studies, as well as observations in human islets, suggest that an improper alignment of glycolysis with TCA-cycle metabolism may underlie perturbed GSIS. As such it may play a role in the pathogenesis of human T2D. Further work is required to establish a casual relationship of these phenomena.

Figure 6 Cartoon of the suggested involvement of HIF-1α in β-cell stimulus–secretion coupling

The cartoons show the proposed different modes of metabolic regulation in the glucose-responsive and -unresponsive clonal β-cell lines involving HIF-1α. AKGA, α-ketoglutarate; HRE, hypoxia-response element; LDH, lactate dehydrogenase; M, mitochondrion; N, nucleus; PHD, proline hydroxylase; PDH, pyruvate dehydrogenase; PDK1, PDH kinase 1; OXPHOS, oxidative phosphorylation; Suc, succinate; VHL, von Hippel–Lindau factor.

AUTHOR CONTRIBUTION

Peter Spégel performed metabolomics analyses and wrote the first draft of the manuscript. Siri Malmgren performed insulin secretion and oxygen consumption measurements. Vladimir Sharoyko performed glucose uptake measurement, provided feedback on results and the manuscript and assisted in metabolite extraction. Thomas Koeck determined HIF-1α stabilization, performed oxygen consumption measurements, and provided feedback on results and the manuscript. Philip Newsholme provided feedback on results and the manuscript throughout the project. Hindrik Mulder conceived and directed the project and finalized the paper prior to publication.

FUNDING

This work was supported by the Swedish Research Council [grant number 14196-06-3], the Crafoord Foundation, the European Foundation for the Study of Diabetes, Lars Hierta, Fredrik and Ingrid Thuring, Åke Wiberg, Albert Påhlsson, O.E. and Edla Johansson Foundations, Knut and Alice Wallenberg Foundation, the Royal Physiographic Society, Lund University Diabetes Centre, and the Faculty of Medicine at Lund University. Support from the Inga and John Hain Foundation to P.S. is acknowledged.

Abbreviations: BCA, bicinchoninic acid; BMI, body mass index; ECL, enhanced chemiluminescence; GC/MS, gas chromatography/mass spectrometry; GK, Goto–Kakizaki; GSIS, glucose-stimulated insulin secretion; HbA1c, glycated haemoglobin; HBSS, Hepes-balanced salt solution; HIF-1α, hypoxia-inducible factor 1α; LDHA, lactate dehydrogenase A; OCR, oxygen consumption rate; SAB, secretion assay buffer; T2D, Type 2 diabetes; TCA, tricarboxylic acid

References

View Abstract