Proton leak exerts stronger control over ATP/ADP in mitochondria from clonal pancreatic β-cells (INS-1E) than in those from rat skeletal muscle, due to the higher proton conductance of INS-1E mitochondria [Affourtit and Brand (2006) Biochem. J. 393, 151–159]. In the present study, we demonstrate that high proton leak manifests itself at the cellular level too: the leak rate (measured as myxothiazol-sensitive, oligomycin-resistant respiration) was nearly four times higher in INS-1E cells than in myoblasts. This relatively high leak activity was decreased more than 30% upon knock-down of UCP2 (uncoupling protein-2) by RNAi (RNA interference). The high contribution of UCP2 to leak suggests that proton conductance through UCP2 accounts for approx. 20% of INS-1E respiration. UCP2 knock-down enhanced GSIS (glucose-stimulated insulin secretion), consistent with a role for UCP2 in β-cell physiology. We propose that the high mitochondrial proton leak in β-cells is a mechanism which amplifies the effect of physiological UCP2 regulators on cytoplasmic ATP/ADP and hence on insulin secretion.
Type 2 diabetes is rapidly becoming a global pandemic. Although the development of this metabolic disorder is understood incompletely, it is evident that pancreatic β-cell dysfunction is responsible for its final progression to hyperglycaemia. To appreciate fully how Type 2 diabetes arises, it is essential to obtain a detailed understanding of β-cell function and regulation. When blood glucose levels rise, β-cells increase their oxidative glucose catabolism, leading to increased mitochondrial protonmotive force and an increased cytoplasmic ATP/ADP ratio. This closes plasma membrane KATP channels, causing plasma membrane depolarization, opening of voltage-sensitive calcium channels, calcium influx and insulin secretion . This canonical model of GSIS (glucose-stimulated insulin secretion) relies heavily on efficient mitochondrial energy transduction, since it appears essential that glucose oxidation is coupled tightly to ATP synthesis.
In general, oxidative phosphorylation is coupled incompletely due to proton leakage across the mitochondrial inner membrane. Proton leak in isolated mitochondria is mainly constitutive and depends on the amount of the adenine nucleotide translocase . Importantly, certain proton leak activity is inducible, which provides a mechanism to regulate the efficiency of energy transduction. Inducible proton leak is mostly accounted for by UCPs (uncoupling proteins). The archetypal uncoupling protein, UCP1, catalyses mitochondrial proton conductance in mammalian brown adipose tissue, where it plays the central role in adaptive thermogenesis. The UCP1 orthologues UCP2 and UCP3 mediate mitochondrial proton leak too, but only when activated by (e.g.) superoxide or reactive oxygen species derivatives . The physiological functions of UCP2 and UCP3 remain uncertain  and are subject to ongoing debate .
Mitochondria from cultured β-cells exhibit GDP-sensitive proton conductance that is stimulated by superoxide . Since UCP2 mRNA is expressed in pancreatic islets, this suggests strongly that β-cells contain a functional UCP2 that is activated by reactive oxygen species . Given its proton-conducting function, it is thought that UCP2 activity attenuates the response of the protonmotive force and ATP/ADP to glucose, and thus impairs GSIS. The most persuasive support for this notion comes from genetic knockout studies, which reveal that pancreatic islets from Ucp2-ablated mice exhibit higher GSIS rates than wild-type controls . Moreover, insulin secretion is improved when endogenous superoxide is lowered using a superoxide dismutase mimetic. Importantly, this effect is observed exclusively in wild-type animals, which demonstrates that UCP2 diminishes GSIS when activated by superoxide . The ability of UCP2 to modulate insulin release is not undisputed: β-cell-specific overexpression of human UCP2 in mice and doxycycline-inducible overexpression of the protein in cultured β-cells did not appear to affect GSIS .
In many cell types, e.g. hepatocytes , thymocytes  and neuronal cells , a significant proportion (20–25%) of cellular respiratory activity is used to drive ‘futile’ mitochondrial proton leak. Despite the potentially pathological effect of UCP2 activity and uncoupling in pancreatic β-cells, little is known about the magnitude and importance of proton leak in this cell type. We recently predicted that proton leak activity may be relatively high in β-cells, because proton conductance is considerably higher in mitochondria isolated from a well-established clonal β-cell line (INS-1E, see ) than in those isolated from rat skeletal muscle. As a consequence of its high rate, proton leak exerts a relatively high control over the Δψ (mitochondrial membrane potential) and ATP/ADP ratio in INS-1E mitochondria . The nature of this relatively high proton conductance is unclear, however, and it is uncertain whether the high leak in INS-1E mitochondria is of physiological relevance or merely represents a mitochondrial isolation artefact.
In the present study, we show that, similarly to the situation in isolated mitochondria, proton leak rates are significantly higher in INS-1E cells than in myoblasts. Myxothiazol-sensitive, oligomycin-resistant respiratory activity (a good approximation of leak) is roughly 4-fold higher in INS-1E than in muscle cells. RNAi (RNA interference) experiments demonstrate that over 30% of the leak activity in INS-1E cells is accounted for by UCP2. We estimate that roughly 20% of INS-1E respiration is caused by proton cycling through UCP2. In agreement with the loss-of-function studies in mice, GSIS is enhanced upon UCP2 knock-down in INS-1E cells.
Cell culture and mitochondrial isolation
INS-1E cells were grown as described in  in a medium containing 11 mM glucose; antibiotics were omitted in RNAi experiments. C2C12 myoblasts were grown at 37 °C under 5% CO2 in Dulbecco's modified Eagle's medium (Sigma–Aldrich, Gillingham, U.K.) supplemented with 10% (v/v) FCS (foetal calf serum) and 4 mM glutamine. Mitochondria were prepared as described in . INS-1E cells and myoblasts harvested from 500 cm2 trays (Nunclon; 80–90% confluence) yielded typically ∼10 and 3 mg of mitochondrial protein per tray respectively. The comparatively low yield from myoblasts is due to the relatively high mechanical force required to rupture cells of this type. Importantly, mitochondria from INS-1E cells and myoblasts were well coupled, with respiratory control ratios during succinate oxidation of ∼3 and 4 respectively.
UCP2 knock-down and detection
siRNA (small interfering RNA) oligonucleotides (Ambion, Huntingdon, U.K.) were targeted at rat Ucp2 exons 3, 5 and 8 (Table 1). Their ability to lower UCP2 levels was assessed in cells seeded in 175 cm2 flasks (BD Falcon™) and grown to ∼50% confluence. Following complex formation, siRNA and the transfection agent Lipofectamine™ 2000 (Invitrogen, Paisley, Renfrewshire, Scotland, U.K.) were added at final concentrations of 50 nM and 1.7 μg/ml respectively. To identify potential non-specific effects, cells were transfected in parallel with scrambled siRNA (Silencer® Negative Control 1; Ambion). After 3 days' further growth, transfected cells were trypsinized; mitochondria were then isolated, resuspended and stored at −80 °C in gel-loading buffer [10% (w/v) SDS, 250 mM Tris/HCl, pH 6.8, 5 mM EDTA, 50% (v/v) glycerol, 5% (v/v) 2-mercaptoethanol and 0.05% (w/v) Bromophenol Blue]. To detect UCP2, 20 μg of mitochondrial protein was separated on an SDS/12% polyacrylamide gel and transferred to nitrocellulose. The membrane was blocked for 2 h at room temperature (21 °C) in Tris-buffered saline containing 0.1% (v/v) Tween 20 and 5% (w/v) dried skimmed milk (Marvel) and then probed overnight at 4 °C with goat anti-UCP2 1° antibodies (Santa Cruz Biotechnology, Santa Cruz, CA, U.S.A.) diluted to 0.2 μg/ml in this blocking buffer. Subsequently, the membrane was incubated for 1 h at room temperature with peroxidase-conjugated 2° antibodies (Pierce Biotechnology, Rockford, IL, U.S.A.) diluted to 0.1 μg/ml in blocking buffer and then cross-reacting proteins were visualized by ECL® (enhanced chemiluminescence) (Amersham Biosciences, Little Chalfont, Bucks., U.K.). Finally, the membrane was stained with GelCode® Blue reagent (Pierce Biotechnology) to confirm equal protein loading. Protein was quantified by densitometry using ImageJ software (http://rsb.info.nih.gov/ij/).
Mitochondrial proton leak
O2-uptake rates and Δψ were measured simultaneously as described in . Mitochondria were incubated at ∼0.9 mg/ml with 350 ng/ml nigericin to collapse ΔpH, 4 μM rotenone to inhibit complex I and 0.9 nmol of oligomycin/mg of protein to inhibit phosphorylation. Succinate (4 mM) was added to initiate state 4 respiration, which was titrated with malonate (up to 0.4 mM) causing stepwise inhibition of respiratory chain activity and decrease in Δψ. Expression of the successive O2-uptake rates as a function of the concomitant Δψ values revealed the kinetic response of proton leak to Δψ.
Non-transfected INS-1E cells and myoblasts, seeded in 175 cm2 flasks, were grown to ∼80% confluence. In RNAi experiments, cells were grown to roughly 50% confluence, transfected as described above and then grown for a further 3 days. Cells were then trypsinized, resuspended in their respective growth media at roughly 107 cells/ml and stored on ice. Cell aliquots (600 μl) were put in glass vessels, flushed with 5% CO2/95% air, sealed, incubated for 30 min in a shaking (100 cycles/min) water bath at 37 °C, and then transferred to an oxygraph. Respiratory rates were calculated from the initial steady parts of the O2-uptake traces and expressed per 106 viable cells. Experiments were also performed with 7 μM oligomycin to approximate proton leak and 9 μM myxothiazol to determine non-mitochondrial O2 uptake. Inhibitors were dissolved in DMSO (final concentration ≤0.2%) and added just before flushing with CO2/air. Both inhibitor concentrations gave maximum effects as checked by doubling concentrations.
Cells were seeded on to 24-well plates (BD Falcon™) at 3×105 cells/well, grown overnight and then transfected with UCP2-targeted (exon 8) or scrambled siRNA. After 3 days' further growth, transfected cells were starved for 2 h in growth medium that lacked glucose and pyruvate and contained only 1% FCS. Cells were washed twice with a glucose-free KRBH [Krebs–Ringer bicarbonate Hepes buffer; 135 mM NaCl, 3.6 mM KCl, 10 mM Hepes, pH 7.4, 0.5 mM MgCl2, 1.5 mM CaCl2, 5 mM NaHCO3, 0.5 mM NaH2PO4 and 0.1% (w/v) BSA] and were then incubated in this buffer for 30 min in a shaking (30 cycles/min) water bath at 37 °C. Next, the medium was replaced with KRBH containing 2, 5 or 30 mM glucose, or 2 mM glucose +30 mM KCl. After a further 30 min, the medium was collected and centrifuged to pellet any detached cells. Supernatants were assayed for insulin by ELISA (Mercodia, Uppsala, Sweden), using mouse insulin as a standard.
Mean differences between experimental systems and conditions were tested for statistical significance by Student's t test.
Higher proton leak in INS-1E cells than myoblasts
INS-1E mitochondria exhibit proton conductance that is significantly higher than that observed in mitochondria from rat skeletal muscle . However, it may be that this difference solely indicates a discrepancy between cell and tissue mitochondria or merely reflects a mitochondrial isolation artefact. As shown by the results in Figure 1, both these concerns are ungrounded. Figure 1(A) demonstrates that proton leak rate in myoblast mitochondria is substantially lower at any measured Δψ than the equivalent rate in INS-1E mitochondria. The considerable leak difference between mitochondria also manifests itself when leak is measured in intact cells. The proton leak rate in cells was estimated as the intracellular mitochondrial respiration rate (defined as respiration sensitive to the electron transport chain inhibitor myxothiazol) that was not sensitive to the ATP synthase inhibitor oligomycin. This method overestimates the leak rate a little because oligomycin causes a small increase in Δψ that slightly increases leak, but based on experiments with hepatocytes  the overestimate is not expected to be large. Although total INS-1E O2 uptake is only approximately half of that observed in myoblasts, the absolute oligomycin-resistant rate is nearly four times higher in INS-1E cells, and relative to the rate without oligomycin, the rate with oligomycin is seven times higher in INS-1E cells (Figure 1B). These cellular differences are greater than predicted from the mitochondrial data, which illustrates the relative importance of respiratory experiments in intact cells.
Higher proton leak in INS-1E cells than myoblasts
From Figure 1(B), the ratio of O2-uptake rates with and without oligomycin can be calculated. This ratio provides an estimation of the proportion of respiratory activity that is used to drive ‘futile’ mitochondrial proton leak in INS-1E (0.76±0.06) and muscle cells (0.11±0.02). Thus 75% of respiration would appear to drive proton leak in INS-1E cells, while only 10% of respiration in myoblasts is used for this purpose. Keeping in mind that the in vitro total respiratory rates used to calculate these proportions are not necessarily representative of those prevalent in vivo, the values obtained may be compared with similar ratios reported in the literature. The equivalent values are ∼25% in hepatocytes  and ∼20% in thymocytes  and neuronal cells . The high INS-1E value disagrees with a value of 10% in INS-1 cells , perhaps due to extensive differences in experimental conditions. Alternatively, this discrepancy could be related to the different origins of the respective cell lines. INS-1 is an insulin-secreting line that was isolated from a radiation-induced rat insulinoma . Because of their non-clonal nature, INS-1 cells exhibit limited stability when passaged, which has been suggested as a possible explanation for the large variability in published results . INS-1E, on the other hand, is a clonal line derived from the parental INS-1 cells , has been characterized in detail and represents a significant improvement in terms of β-cell differentiation and stability .
UCP2 contributes to high INS-1E proton leak
Given the pronounced effect of UCP2 ablation on GSIS in pancreatic islets , we reasoned that the relatively high proton leak activity observed in INS-1E cells (Figure 1B) could be due to UCP2 activity. We used RNAi to reduce UCP2 levels, and thus test this hypothesis. We have routinely used a polyclonal antibody to detect UCP2 in mouse tissues, and found no UCP2 band in samples from UCP2-knockout mice (results not shown). Figure 2(A) shows that this polyclonal antibody cross-reacts with a single protein of ∼32 kDa in samples obtained from both non-transfected cells (lane 1) and those transfected with scrambled siRNA (lane 5). The intensity of this band is reduced dramatically in samples from cells that have been transfected separately with three different siRNA oligonucleotides targeted at independent Ucp2 exons (lanes 2–4). Densitometric comparison confirmed equal loading of total mitochondrial protein (Figures 2B and 2D), and revealed that UCP2 levels are reduced substantially within 72 h of transfection (Figure 2C). The observed densities upon UCP2 knock-down are roughly 15% (exons 5 and 3) and 10% (exon 8) of the scrambled control.
O2-consumption rates of cells transfected with scrambled siRNA were 10.6±2.1 in the absence and 7.7±1.5 nmol of O/min per 106 cells in the presence of oligomycin. These activities are comparable with those observed in non-transfected INS-1E cells (Figure 1B). The overall respiratory activity of cells transfected with Ucp2-targeted siRNA is on average nearly 13% lower than that of cells transfected with scrambled siRNA (Figure 3A). This difference is statistically significant (P=0.02) and, importantly, it arises irrespective of the particular Ucp2 exon targeted. Similarly, oligomycin-insensitive respiratory activity is lowered, on average by more than 30% (P=1×10−5), when UCP2 levels are diminished by siRNA targeted at three separate exons (Figure 3B). As a result, the fraction of overall respiration that is oligomycin-resistant is reduced significantly (P=0.01) upon UCP2 knock-down (Figure 3C). This suggests strongly that UCP2 contributes substantially to the proton leak rate of non-transfected INS-1E cells.
UCP2 knock-down lowers proton leak
In this context, it is anticipated that Δψ increases upon UCP2 knock-down. This will not only lead to an underestimation of UCP2's contribution to proton leak (see the Discussion section), but may also result in increased reactive oxygen species levels. It is possible that this in turn activates additional uncoupling mechanisms. However, the nature of UCP2-independent proton leak is unclear at present.
UCP2 knock-down enhances insulin secretion
In the present experiments, insulin secretion rates in INS-1E cells were relatively high at 2 mM glucose. However, these rates were increased significantly when the glucose level is raised to 30 mM, or cells were incubated additionally in the presence of 30 mM KCl (Figure 4A). Since INS-1E cells are expected to respond to such stimuli in this manner , we deemed our assay suitable to assess the effect of UCP2 knock-down on GSIS. In cells transfected with scrambled siRNA, insulin secretion was similar at 2 and 5 mM glucose (Figure 4A), but while UCP2 knock-down does not affect secretion at 2 mM glucose, it enhanced insulin release at 5 mM glucose. This indicates that GSIS is improved, which agrees with observations made in islets from Ucp2-knockout mice . Upon UCP2 knock-down, insulin secretion was also increased at 30 mM glucose but, importantly, KCl-induced secretion, which does not rely on mitochondrial energy metabolism, was not affected (Figure 4A). From Figure 4(B), it can be seen that cell number and cellular insulin content are not altered significantly after UCP2 down-regulation, although the average total insulin level appears somewhat increased.
UCP2 knock-down enhances GSIS
Because the cytoplasmic ATP/ADP of pancreatic cells is governed predominantly by ATP supply , these cells are interesting from a bioenergetic perspective. Glucose-induced increases in ATP/ADP observed in pancreatic cells  are not seen in most other cell types, where this ratio is mainly controlled by ATP demand . Based on a metabolic control analysis of oxidative phosphorylation in muscle and INS-1E mitochondria, we suggested recently that this difference in control is partly due to mitochondrial peculiarities . In particular, proton leak rate was relatively high in INS-1E mitochondria and, consequently, exerted relatively high control over Δψ and the external ATP/ADP ratio. The results reported in the present paper demonstrate that a relatively high proton conductance does not merely represent a mitochondrial isolation artefact, but is a characteristic feature of intact INS-1E cells (Figure 1B). The estimated proportion of overall respiratory activity used to drive ‘futile’ proton leak in INS-1E cells is much larger than that used in hepatocytes , thymocytes  or neuronal cells .
Our results establish that a large part of the high leak rate in INS-1E cells is accounted for by UCP2, as oligomycin-resistant O2 consumption is decreased by ∼30% when the protein is knocked down using RNAi (Figure 3B). Figure 3(A) indicates that UCP2 also contributes to the total respiratory activity. A contribution of 13% is an underestimate, however, because UCP2 knock-down will raise Δψ and cause a compensatory increase in proton leak. A better estimate comes from multiplying the slightly underestimated contribution of UCP2 to the oligomycin-insensitive respiration (30%) by the slightly overestimated proportion of overall respiration that drives proton leak (75%) to give a value of approx. 20% of the cell respiration rate caused by proton cycling through UCP2 in cells. This value is calculated assuming that RNAi removes all UCP2, which will cause a further small underestimation of its contribution. Nonetheless, 20% is high compared with the calculated 10% in thymocytes: ∼20% of thymocyte respiration is caused by proton cycling  and ∼50% of this is lost in thymocytes from UCP2-knockout mice .
The relatively high contribution of UCP2 to INS-1E respiration (Figure 3), its attenuating effect on GSIS (Figure 4A) and the significant control of proton leak over ATP/ADP  support the view that UCP2 may play an important regulatory role in β-cell physiology. Our results indicate that UCP2 is most likely active under pathological and physiological conditions with its activation state being dependent on fluctuating levels of effectors such as superoxide . In other words, UCP2 activation is unlikely to be just a pathological side effect of a damage limitation mechanism which is induced during high-fat feeding or hyperglycaemia and which causes cell dysfunction [7,8,19]. We suggest that the relatively high mitochondrial proton leak in cells is a mechanism that amplifies the effect of physiological regulators of UCP2 proton conductance on the cytoplasmic ATP/ADP ratio and hence on GSIS.
The notion that UCP2 regulates GSIS is corroborated by a steadily increasing body of evidence. Initial studies showed that adenoviral overexpression of UCP2 leads to attenuated GSIS in rats  and insulinoma cells . Despite the fact that overexpression of UCPs tends to cause non-specific proton conductance , such observations demonstrate that GSIS is impaired when oxidative phosphorylation becomes less efficient. More importantly, investigations involving Ucp2-ablated mice showed convincingly that UCP2 is a negative modulator of GSIS under conditions that are of (patho)physiological relevance [7,8,19,22]. Our results (Figure 4A) agree with these gene knockout studies: lowering UCP2 levels in INS-1E cells through RNAi results in increased GSIS. Results obtained with INS-1E cells are inevitably of less physiological relevance than those obtained using islets. Arguably, however, the interpretation of experiments with Ucp2-knockout mice is relatively complicated, since the absence of the Ucp2 gene may have caused potentially complex developmental adaptations. Additionally, the observed altered islet cell behaviour may be an indirect result of the absence of Ucp2 in other tissues such as the brain. Because our experimental model does not suffer from such complications, the results reported in the present paper demonstrate that GSIS can be modulated acutely by UCP2 activity. Such an acute regulation is in line with observed immediate effects on GSIS of a superoxide dismutase mimetic  and a pharmacological UCP2 inhibitor . It is also worth noting that acute short-term inhibition of UCP2 expression by antisense oligonucleotides has recently been shown to ameliorate the hyperglycaemic syndrome in two different animal models of Type 2 diabetes . Interestingly, this metabolic improvement appears to be due to a combined effect of UCP2 knock-down on insulin secretion by pancreatic cells and insulin action on peripheral tissues.
Despite the mounting evidence for a GSIS-modulatory role, it remains uncertain whether UCP2 indeed affects insulin secretion through uncoupling ATP synthesis from mitochondrial electron transfer. Previously reported effects of UCP2 on ATP/ADP (see e.g. [7,23]), Δψ [8,22] and O2 consumption  favour such an uncoupling function, but are indirect measures providing circumstantial support only. In fact, the strongest direct indication that UCP2 uncouples oxidative phosphorylation in situ in any mammalian cell comes from kinetic experiments with thymocytes . At any given Δψ, thymocytes from Ucp2-knockout mice exhibit lower non-phosphorylating (i.e. state 4) respiratory activity than wild-type cells. The effect of UCP2 knock-down on oligomycin-resistant (i.e. state 4) O2 uptake in INS-1E cells (Figure 3) is in close agreement with the kinetic measurements in thymocytes and provides direct evidence for an uncoupling function of UCP2 in cells.
The results reported in the present paper validate predictions of relatively high proton leak activity in INS-1E cells and demonstrate that UCP2 contributes substantially to this activity. They show, furthermore, that UCP2 modulates insulin secretion in INS-1E cells in a fashion that is similar, qualitatively, to that observed in pancreatic islets. Our work thus highlights that INS-1E cells are a convenient and appropriate model to further clarify the role of UCP2 in islet cell physiology.
INS-1E cells were donated by Dr P. Maechler and Dr C. Wollheim (Department of Internal Medicine, University Medical Center, Geneva, Switzerland). This work was supported by the U.K. MRC (Medical Research Council).