Reactive oxygen species and uncoupling protein 2 in pancreatic β-cell function

What Types of ROS May Function as Cellular Signals?

Mitochondrial respiration and various oxidoreductases are the major source of cellular ROS. ·O₂⁻ is a very reactive molecule, but it can be converted to less reactive H₂O₂ by superoxide dismutase (SOD) isoenzymes, and then to oxygen and water mainly by catalase (CAT), glutathione peroxidases (GPX) and peroxiredoxin (PXR). In addition, ·O₂⁻ may form hydroxyl radical (HO·) by Fenton's reaction or react with nitric oxide (NO) to generate peroxynitrite (ONOO⁻) under certain instances, such as inflammatory conditions (figure 1). HO· and ONOO⁻ are very toxic and can cause oxidative damage. Unlike other ROS, H₂O₂ is a small, stable, uncharged, freely diffusible molecule that can be rapidly synthesized and destroyed in response to external stimuli [18]. By this definition, H₂O₂ arguably meets the criteria for an intracellular messenger. Indeed, a growing body of evidence supports the notion that H₂O₂ is a ubiquitous intracellular messenger [4,9]. In the case of pancreatic β-cells, they are equipped with moderate, but physiologically sufficient, catalytic capacities for conversion of ·O₂⁻ into H₂O₂ in cytoplasm and mitochondria [19]. However, levels of the H₂O₂-inactivating enzymes GPX and CAT are extremely low in β-cells, comprising for example only 1% of its expression level in the liver [20]. This apparent imbalance between O₂⁻ and H₂O₂-inactivating enzymes in β-cells potentially favours H₂O₂ accumulation, thus rendering them sensitive to H₂O₂-mediated signal transduction.

Endogenous Sources of ROS in the Cell

Mitochondria are the main source of ·O₂⁻[21]. In the process of cell respiration there are at least three stages that are associated with increased ·O₂⁻ generation. These include increased substrate supply, decreased ADP concentration and increased intracellular Ca²⁺ concentration [22]. The main sites of ·O₂⁻ generation in mitochondria are the inner mitochondrial membrane: NADH dehydrogenase at complex I, and the interface between ubiquinone and complex III [21,23]. Mitochondrial ROS production has been proposed as a necessary stimulus for GSIS [10].

Another source of ROS in cells is the NADPH oxidase (NOX), which is a multi-subunit protein complex located on the cell membrane [24,25], and probably best known for its role in the immune cell respiratory burst. Activated NOX takes an electron from donor NADPH and translocates it across the cell membrane to an extracellular O₂ molecule, generating O₂⁻·. Pancreatic islets constitutively express multiple NOX isoforms [26] which could conceivably play a role in ROS production by pancreatic β-cells during GSIS [27]. In addition to these well-studied mechanisms for ROS production, our recent in vitro experiments indicate that glucose autoxidation is a powerful source of H₂O₂ generation [13], which may also contribute to GSIS in β-cells. Other endogenous sources of ROS include some CYP enzymes and flavoproteins in endoplasmic reticulum (ER), lipoxygenases, nitric oxide synthetase isoforms and prostaglandin synthase on plasma membrane, various oxidases and flavoproteins in peroxisomes and xanthine oxidase in cytoplasm, etc. However, the source of this ROS and the biochemistry involved in regulating β-cell function is not clear.

Downstream Targets of ROS Signals

Intracellular redox status is very sensitive to changes in ROS generation and antioxidant activity [28]. Small variations of intracellular ROS have been shown to modulate many physiological processes including redox-dependent transcriptional regulation [29,30], ion transport [31], as well as protein phosphorylation [3,18]. For H₂O₂ in particular there is evidence for its activation of a number of intracellular signalling molecules, including protein phosphatases, kinases and transcription factors [32]. Physiological responses to signals such as epidermal growth factor (EGF) and insulin show that a primary target of the H₂O₂ generated is protein tyrosine phosphatase (PTP) [8]. By oxidizing the thiol groups on the catalytic cysteine residues of PTP to sulfenic or sulfinic acid, H₂O₂ inhibits catalytic activity [8]. This inhibition of PTP consequently permits heightened phosphorylation of substrate proteins and signal transduction. In the case of EGF, the inhibition results in enhanced EGF receptor autophosphorylation and subsequent mitogen activated protein kinase (MAPK) activation, which promotes proliferation, adhesion and migration. In the case of insulin, the inhibition of PTP results in augmented phosphorylation of insulin receptor and downstream mediators, leading to increased glucose uptake [9,33]. In pancreatic β-cells, several signal transduction molecules or processes that are associated with GSIS and/or β-cell function are being recognized as downstream targets of ROS, H₂O₂ in particular. These molecules and processes include Ca²⁺-dependent protein phosphatases [34], PTPs [8,35], voltage-gated K⁺ channels [36], Ca²⁺ influx and release [37–40], tumour suppressor phosphatase PTEN [18], c-Jun N-terminal kinase [41], extracellular signal-regulated kinases [42], NF-κB [43] and SIRT1 deacetylase [44,45]. However, the biochemistry by which H₂O₂ mediates GSIS is still under investigation.

A ROS Signal Contributes to GSIS

Insulin secretion is subject to control by nutrients as well as by hormonal, neural (and pharmacological) factors. Among these, glucose is by far the most important regulator of the machinery of insulin secretion [46]. It has been well documented that glycolytic and oxidative events leading to accelerated ATP generation are key transduction phenomena in β-cell signalling. However, the generation of ROS is also coupled to this glycolytic and respiratory metabolism in β-cells [12]. As ROS have emerged as physiological mediators of many cellular responses [47,48], this raises the possibility that these molecules could serve a signalling function in glucose responsiveness. To support ROS as metabolic signals in regulating GSIS, our recent studies [13] indicated that (i) glucose stimulates H₂O₂ generation and alters intracellular redox status; (ii) an increase in intracellular H₂O₂ increases insulin secretion; (iii) suppression of glucose-induced H₂O₂ accumulation by antioxidants impedes GSIS. These findings suggest that H₂O₂ derived from glucose metabolism is one of the metabolic signals for insulin secretion. This mechanism is strongly supported by Leloup et al., who showed that mitochondrial ROS are obligatory signals for GSIS [10]. Again, the exact mechanism by which ROS regulates GSIS, such as the identity of the downstream target(s) of ROS, is unclear.

UCP2 and Oxidative Stress

The ·O₂⁻ production from the mitochondrial matrix is very sensitive to the proton motive force [49], so mild uncoupling can substantially decrease mitochondria-derived ROS and is believed to aid in preventing oxidative damage [49–51]. UCP2 is a widely expressed mitochondrial inner membrane carrier protein that was discovered through its sequence homology to the brown fat-specific UCP1. UCP1 dissipates caloric energy as heat by uncoupling mitochondrial respiration from ATP production [14]. Although the biochemical functions of UCP2 are still much debated [52], there is strong evidence that UCP2 is not a physiologically relevant ‘uncoupling protein’ in the manner of UCP1 and does not contribute to adaptive thermogenesis [17,53]. Instead, accumulating evidence supports the idea that UCP2 participates in the control of mitochondria-derived ROS [17,51]. For example, the genetic absence of UCP2 results in increased ROS production in macrophages [54,55], while acute overexpression of UCP2 in vitro and in vivo has been shown to protect against overt oxidative damage [56,57]. More recently, a role has been presented for UCP2 in metabolic sensing through mitochondrial respiration in hypothalamic neuronal populations [58]. In addition, as with most antioxidant responses, the transcription of the Ucp2 gene itself is highly inducible under conditions of oxidative stress. For example, agents such as H₂O₂[56], lipopolysaccharide [59], TNFα[60], free fatty acids [61], irradiation [62] as well as high-fat diet challenge [63,64] have all been shown to increase UCP2 expression in vitro or in vivo. It has also been reported to increase the proton conductance of the mitochondrial inner membrane when UCP2 is activated by ·O₂⁻[65,66] and/or free radical-derived alkenals such as 4-hydroxy-2-nonenal [17,67]. Thus, there is good evidence that (i) a major physiological function of UCP2 is to attenuate mitochondrial production of ROS, and (ii) activation and/or induction of UCP2 may function as an adaptive response to oxidative stress, whereas the UCP2-mediated feedback regulation on ROS generation may be considered as a compensatory mechanism alleviating oxidative stress [59].

Pancreatic β-cell Phenotype of Ucp2-null Mice

The signalling role of ROS in GSIS coupled with the negative regulatory effect of UCP2 on mitochondria-derived ROS supports a hypothesis that UCP2 is an endogenous suppressor of insulin secretion [68]. Consistent with this hypothesis, overexpression of UCP2 in isolated β-cells has been reported to inhibit GSIS [69–71] and short-term ‘knockdown’ of UCP2 in mouse islets or inhibition of UCP2 activity with a small molecule has been reported to acutely increase GSIS [72,73].

Given the overwhelming evidence that UCP2 is a physiologically relevant negative regulator of ROS production [17,51,54,55], the chronic absence of this protein as achieved by targeted deletion methods has the potential to lead to persistent ROS accumulation and an adaptive antioxidant response, in particular in those tissues with high basal UCP2 expression, including pancreatic β-cells. In our recent study, we provided direct in vivo evidence that the absence of UCP2 in mice results in significant oxidative stress [74]. In Ucp2−/− islets, a decreased GSH and/or elevated GSSG as well as elevated levels of nitrotyrosine were observed, suggesting persistent local oxidative stress.

Oxidative stress is a common denominator amongst the various mechanisms proposed for β-cell dysfunction and the progression to frank diabetes [75–77]. In most tissues, cells exhibit an early adaptive response to oxidative stress that stimulates the cellular antioxidant system to protect them from oxidative damage. Such an acute compensatory response is generally appropriate to protect against ROS toxicity [2,78]. However, over time the combination of chronic oxidative stress coupled with continuously elevated antioxidants perturbs normal homeostasis, and may include interfering with ROS that may be acutely generated to serve a signalling function. As the transient generation of ROS in β-cells in response to glucose has been proposed to serve as such a signal for insulin secretion [6,10,12,13], this scenario may apply to GSIS in Ucp2−/− mice. For example, under low-glucose conditions H₂O₂ levels were already elevated in islets from Ucp2−/− mice, and this was accompanied by increases of several antioxidant enzymes, including the H₂O₂-scavenging enzymes GPX and CAT [74]. However, in response to a glucose challenge, the net increase in H₂O₂ production seen in the Ucp2−/− islets was substantially lower than what we typically observe in wild-type mice, and was also accompanied by a reduced GSIS in Ucp2−/− islets [74]. Therefore, one hypothesis stemming from these findings is that the adaptive response to chronic oxidative stress caused by the absence of UCP2 may have interfered with the ‘beneficial’ aspects of ROS as a glucose-dependent signal for insulin secretion. Direct testing of this hypothesis will require the future generation and validation of additional models, such as mice lacking both Nrf2 and Ucp2.

Based upon our findings that disruption of the Ucp2 gene causes persistent oxidative stress in general and impairs β-cell function, it raises the question as to whether inhibiting expression or activity of UCP2 is an appropriate therapeutic approach for type 2 diabetes to improve GSIS. Considering the other negative consequences resulting from the absence of UCP2, including increased development of atherosclerotic plaques [79,80], neurodegeneration [81] and heightened propensity for colon tumours [82], caution is warranted regarding this approach. Certainly the chronic absence of UCP2 likely represents a significantly different physiology from any acute moment-to-moment manipulation of its level or activity. Therefore, given the apparent paradoxical roles for UCP2 in β-cell function that now exist, a better understanding of the role of UCP2 in the function of pancreatic β-cells and pathogenesis of diabetes is needed.

A Cautionary Tale Regarding Phenotypes of ‘Knock-out' Mice

The first report describing the generation of Ucp2−/− mice from Arsenijevic et al. described a phenotype of exaggerated ROS and phagocytosis in macrophages [54]. Another report of Ucp2−/− mice showed dramatically enhanced GSIS in Ucp2−/− mice [68]. This finding in β-cells of Ucp2−/− mice was replicated by Collins and Corkey [74]. Of note, all studies on Ucp2−/− mice during that time were conducted with mice that were a C57BL/6J (B6) and 129 mixed background. However, particularly because immune system characteristics of mice are highly influenced by genetic background, the mice were backcrossed onto one of three strains of inbred mice (B6, A/J or 129/SvImJ) for 8–20 generations. Although the macrophage inflammatory phenotype of the mice was unchanged [55] we made the unexpected observation that instead of heightened or improved β-cell function, there was uniformly a significant decrease in GSIS from islets of Ucp2−/− mice, irrespective of background strain [74] (figure 2).

This diametrically opposite result for the pancreatic phenotype from Ucp2−/− mice of 129/B6 mixed background vs. the congenic lines suggests that genetic background strongly affected the phenotype and the biological interpretation. Concerns regarding the contribution of genetic background to the interpretation of phenotypes in targeted knock-out studies are not new. Genetic background and the confounding effects of genes that flank the area of the targeted allele in genetically manipulated mice have been reviewed and commented upon by others [83–86]. As discussed [83,85,86], for ‘knockout’ mice generated by the approach using embryonic stem (ES) cells derived from 129 substrains crossed with B6, selection for the null allele derived from 129 also includes a large number of flanking genes surrounding the targeted allele, while littermates with the wild-type allele possess the equivalent regions from B6 (figure 3). In most cases, this genetic bias is of limited concern. However, when the phenotype ascribed to the targeted disruption is already vastly different between the two inbred strains (see below), it is prudent to consider the possibility of a confounding effect of genetic background. In such cases backcrossing to several inbred strains significantly reduces the flanking area that is linked to the selected targeted allele, in addition to achieving homozygosity at all other loci [83–86].

In the case of B6 mice it has already been established by others that this strain exhibits a remarkably lower glucose tolerance and defective GSIS from islets compared with other inbred strains [87–90]. Note, however, that for isolated islets from 129 and B6 mice obtained from Jackson Laboratories and compared side by side, there is a 15-fold higher GSIS in 129 mice than B6 mice, a slightly higher basal level of secretion from B6, but no differences in total islet insulin content [74] (figure 4). This striking difference between 129 and B6 mice, coupled with the reversal of phenotype in Ucp2−/− mice of 129/B6 mix generated vs. in the B6, 129 or A/J backgrounds, together suggests that we cannot dismiss genetic background as a potential contributor to the widely reported heightened secretion observed in Ucp2−/− mice of mixed strain parentage [68,73]. Indeed, we were able to independently replicate the originally reported phenotype of increased GSIS using our own 129/B6 Ucp2−/− mice [74]. Therefore, given this dramatic difference in GSIS between 129 and B6 islets, it raises the distinct possibility that a locus that resides in the vicinity of the Ucp2 gene—particularly in non-backcrossed animals—contributes to the robust GSIS of the 129 mouse. It should be noted from the Pi et al. data [74] that, despite the uniform degree of impaired secretion that results from the loss of UCP2 within each of these individually backcrossed strains, the significant differences in amplitude of GSIS are still evident among the strains.