Mitochondrial uncoupling prevents cold-induced oxidative stress: a case study using UCP1 knockout mice

The idea of a negative impact of high metabolic rate on longevity was first formulated almost a century ago by Raymond Pearl (Pearl, 1928) in his ‘rate of living theory’. Almost 30 years later, in his ‘free radical theory of ageing’, Denham Harman proposed, as an underlying mechanism, the fact that aerobic respiration leads to the inevitable by-production of damaging reactive oxygen species (ROS)/free radicals (Harman, 1956). Ageing could thus result from the accumulation of oxidative damage, caused by the imbalance between ROS production and antioxidant defences (i.e. oxidative stress), with the rate of ROS production being potentially coupled to whole-organism oxygen consumption and, in turn, metabolic rate (Beckman and Ames, 1998). The production of mitochondrial ROS has sometimes been assumed to be a fixed percentage of total oxygen consumption, falling somewhere between 0.1 and 4% according to in vitro experiments (Golden and Melov, 2001; Nicholls et al., 2002), but whether the same values are found under in vivo circumstances remains to be demonstrated. Nevertheless, one common prediction of the free radical theory of ageing has been that an increase in metabolic rate (i.e. oxygen consumption) should lead to an increase in mitochondrial ROS production and concomitant rate of ageing (Beckman and Ames, 1998).

However, recent evidence, mostly coming from our more accurate understanding of mitochondrial functioning, argues against a trivial, monotonic relationship between oxygen consumption and ROS production (reviewed in Murphy, 2009; Lambert and Brand, 2009; Speakman and Selman, 2011). During mitochondrial respiration, some electrons can escape the electron transport chain and react directly with molecular oxygen to form ROS. The energy associated with the electron flow through the respiratory chain is used to pump protons against their electrochemical gradient across the mitochondrial inner-membrane, and the backflow of protons through the Fo/F1 ATP synthase is responsible for the conversion of cellular energy as ATP. Hence, mitochondria couple respiration to ATP synthesis through an electrochemical proton gradient (Divakaruni and Brand, 2011). A growing number of studies show that electron loss from the respiratory chain and concomitant ROS production are highly sensitive to changes in mitochondrial inner-membrane potential, ROS production sharply declining at low membrane potential (Barja, 2007; Murphy, 2009; Mookerjee et al., 2010). Several pathways, including the inducible uncoupling proteins (UCP1 to UCP3) (reviewed in Ricquier and Bouillaud, 2000), might lower the mitochondrial membrane potential by increasing its proton permeability, thereby uncoupling respiration from ATP production and releasing energy as heat. The occurrence of such a mitochondrial uncoupling can then increase oxygen consumption while lowering ROS production, which in turn might lead to a negative association between metabolism, ROS production and rate of ageing, as stated by the ‘uncoupling to survive hypothesis’ (Brand, 2000; see also Speakman et al., 2004).

In such a context, experimental increases of the uncoupling state of mitochondria, either obtained through a pharmacological uncoupling treatment (2,4-dinitrophenol) (Caldeira da Silva et al., 2008) or through the ectopic (i.e. muscular) expression of the uncoupling protein UCP1 (Gates et al., 2007) in mice, have been shown to extend lifespan. Nevertheless, the relevance of mitochondrial uncoupling in the control of ROS production under

normal physiological conditions remains unclear and controversial (Shabalina and Nedergaard, 2011). Some of these controversies could be attributable to the fact that the impact of UCP1 on oxidative stress levels in situations where UCP1 expression/activity is naturally triggered (e.g. chronic cold exposure) has been under-evaluated so far, despite the fact that UCP1 is the only acknowledged UCP with a physiologically relevant uncoupling activity (Shabalina and Nedergaard, 2011). Early evidence suggested that UCP1 ablation had no impact on ROS production and various markers of oxidative stress in brown adipose tissue (BAT) mitochondria (Shabalina et al., 2006). However, recent studies have demonstrated that UCP1 activity can reduce ROS production of BAT mitochondria, at least under in vitro conditions (Dlasková et al., 2010; Oelkrug et al., 2010). Because these studies only focused on BAT mitochondria, we are still lacking vital information about the impact of UCP1 activity on oxidative homeostasis at the organismal scale, knowing that UCP1 seems crucial to survive during long-term cold exposure (Golozoubova et al., 2001).

The main function of UCP1 is to uncouple ATP synthesis from respiration, leading to heat production (i.e. UCP1 was first referred to as thermogenin) (Cannon et al., 1982). UCP1 is the most abundant protein in the BAT, and this tissue has a central role in the progressive substitution of muscular shivering in response to cold exposure by an endogenous production of heat (Fig. 1A), a process referred as to adaptive nonshivering thermogenesis (NST) (Klingenberg, 1990; Cannon and Nedergaard, 2004). In the present study, we used an experimental design where metabolism was increased with or without the triggering of an uncoupled mitochondrial state (UCP1 activity). To do so, we compared the metabolic rate and oxidative stress markers of wild-type (WT) mice with those of mice deficient for UCP1 (UCP1-KO) housed at 26°C or after a four-week exposure to moderate cold (12°C). Cold exposure is known to trigger NST in small mammals, but this response cannot be used by UCP1-KO mice, which instead have to rely mostly on muscular shivering thermogenesis and efficient production of ATP (i.e. coupled respiration) to fuel muscle activity (Golozoubova et al., 2001). Based on the existing literature, WT mice were expected to rely only on UCP1-dependent nonshivering thermogenesis after ~3 weeks of cold acclimation because NST gradually replaces shivering (Fig. 1A) (Cannon and Nedergaard, 2004) whereas UCP1-KO mice were expected to have to maintain high-intensity shivering for the long-term (Fig. 1B) (Golozoubova et al., 2001). This two-by-two experimental design [genotype (WT versus UCP1-KO) and temperature (12°C versus 26°C)] allowed us to determine oxidative damage, oxidative challenge (GSSG/GSH ratio; see Materials and methods for details) and antioxidant capacities across four groups of mice, both at the plasmatic level and in the two main thermogenic tissues (i.e. skeletal muscles and BAT). We predicted that a cold-induced rise in metabolism should be associated with higher oxidative damage in UCP1-KO mice, whereas no such relationship should be observed in WT mice relying on UCP1-dependent nonshivering thermogenesis. These two alternative hypotheses predict strong genotype × temperature interactions for markers of oxidative stress.

Exposure to moderate cold ambient temperature increased mean oxygen consumption () by more than twofold (Fig. 2; temperature, F=1509.8, P<0.001), but independent of mouse genotype (genotype, F=0.23, P=0.64; interaction, F=1.43, P=0.24). Body mass was entered as a covariate in the model to control for the increase in with body mass (body mass, F=15.03, P=0.001).

Our experimental exposure of WT mice and mice deficient for UCP1 to moderate cold revealed strongly significant genotype × temperature interactions (P≤0.007) on markers of oxidative stress measured in the plasma (i.e. d-ROMs) and the skeletal muscles (i.e. proportion of glutathione oxidized and protein carbonyl content) but not in the BAT (Table 1, Figs 3, 4 and 5). These interactions were explained by the increase in the aforementioned markers of oxidative stress in response to moderate cold exposure for UCP1-KO mice (Fig. 3A, Fig. 5B,C). In contrast, WT mice showed no or only a moderate cold-induced increase in oxidative stress restricted to the proportion of glutathione oxidized in skeletal muscles (Fig. 5B). Here, note that the activity of the antioxidant enzyme glutathione reductase (GR) in skeletal muscle was significantly affected both by the genotype (F=14.35, P=0.001) and by the temperature (12°C>26°C; F=26.94, P<0.001), but not by the interaction between these two factors (F=2.18, P=0.149) (see Fig. S1 in supplementary material). Indeed, UCP1-deficient mice presented higher GR activity than WT mice, and for both genotypes the enzymatic activity was higher in the cold. Plasma antioxidant capacity and tissue total glutathione content did not significantly differ according to mouse genotype or temperature (Table 1, Fig. 3B, Fig. 4A, Fig. 5A).

Cold exposure has been previously used to assess how an increased metabolism may impact ageing in rodents (Holloszy and Smith, 1986; Topp et al., 2000; Selman et al., 2002; Kaushik and Kaur, 2003; Venditti et al., 2004; Selman et al., 2008; Vaanholt et al., 2009). Several studies found significant short- to mid-term effects (from 10 h to 3 weeks) of cold challenge on oxidative stress markers, with, for example, increased oxidative damage (Topp et al., 2000; Selman et al., 2002; Kaushik and Kaur, 2003; Venditti et al., 2004) or tissue-specific modifications of antioxidant defences, which globally reflect a situation of oxidative stress (Kaushik and Kaur, 2003; Venditti et al., 2004). However, while always inducing a rise in metabolism, mid- to long-term cold challenge experiments produced contrasting results. Despite higher metabolic rate in the cold, oxidative stress markers, and ultimately individual survival, were not markedly affected by long-term (i.e. throughout adult life) cold exposure in small rodents (Holloszy and Smith, 1986; Selman et al., 2008; Vaanholt et al., 2009). Thermoregulatory mechanisms implicated in the cold response may be, at least partially, responsible for these discrepancies. An underestimated phenomenon is that thermogenesis is primarily achieved through muscular shivering in the hours to days of exposure to cold but is progressively replaced by adaptive nonshivering thermogenesis (Fig. 1A) (Janský, 1973; Klingenspor, 2003; Cannon and Nedergaard, 2004; Ouellet et al., 2012). This latter process is achieved through mitochondrial uncoupling via UCP1 in the brown adipose tissue (extensively reviewed by Cannon and Nedergaard, 2004). Interestingly, longevity is shortened in UCP1-KO mice during prolonged cold exposure, with a median survival of ~13 weeks compared with >24 weeks for WT mice (Golozoubova et al., 2001). Following a period of acclimation at 18°C, mice lacking UCP1 could maintain body temperature and resist cold temperatures through continuous shivering but apparently at a cost for longevity. We confirmed here that a moderate cold exposure (26°C to 12°C) approximately doubles the metabolic rate, but independently of the mice genotype, as previously demonstrated (Golozoubova et al., 2001; Meyer et al., 2010). The longevity effect observed previously (Golozoubova et al., 2001) was probably not mediated by differences in terms of metabolic rate per se, but our results suggest that cold-induced oxidative stress occurs in mice lacking UCP1, which may contribute to explaining the reduced longevity of these mice in the cold.

Because cold-induced increase in metabolism at the whole-organism level is related to the higher activity of a few specific tissues (i.e. muscular shivering thermogenesis and/or BAT nonshivering thermogenesis), the impact of cold-induced high metabolism on the oxidative balance is likely to be tissue dependent (Kaushik and Kaur, 2003). Accordingly, following cold exposure, UCP1-KO mice showed greater levels of oxidative stress in the blood and in skeletal muscles, but not in BAT, compared with WT mice. In the absence of a cold challenge, WT and UCP1-KO mice had similar levels of BAT/muscles total glutathione content, oxidative challenge (proportion of glutathione oxidized) or oxidative damage of proteins. Hence, our results suggest that, in the absence of a cold challenge and concomitant overexpression of UCP1, BAT has no major influence on oxidative stress (as previously suggested by Shabalina et al., 2006). Furthermore, once UCP1 expression is triggered, we found that BAT metabolism activation during NST has neither a local deleterious effect (i.e. on BAT) nor a systemic pro-oxidant deleterious effect (i.e. on the muscles and plasma). This is remarkable given that BAT metabolism is dramatically increased during cold exposure (Cannon and Nedergaard, 2004) and that brown adipocytes contain numerous mitochondria and then have a high oxidative capacity (Ricquier and Bouillaud, 2000). Therefore, even if UCP1 overexpression does not directly reduce oxidative stress following cold exposure, it likely reduces the proportion of ROS generated per unit of oxygen consumed. The induced uncoupling mitochondrial state due to UCP1 activity could be one of these processes and contributes to maintain redox homeostasis in the BAT during thermogenesis. Such a UCP1 secondary effect (i.e. in addition to its thermogenic effect) in brown adipocytes is supported by in vitro experiments (Dlasková et al., 2010; Oelkrug et al., 2010) showing that UCP1 expression reduces ROS production in isolated mitochondria. Our results are also in line with a recent report of beneficial health effects of overexpression of the tumour suppressor Pten in transgenic mice, those health effects being associated with striking hyperactivity of BAT and increased levels of UCP1, which in turn were leading to high metabolic rate but low levels of oxidative damage and lifespan extension (Ortega-Molina et al., 2012).

Although we did not measure shivering per se in the present study, previous studies on cold acclimation between WT and UCP1-KO mice point out that after more than 4 weeks in the cold WT mice are expected to rely only on NST whilst UCP1-KO mice are expected to still rely on shivering (Fig. 1) (Golozoubova et al., 2001; Cannon and Nedergaard, 2004). Therefore, our results suggest that NST and muscular shivering thermogenesis lead to similar cold-induced increases in metabolic rate after 4 weeks of mild cold exposure. The protective effect of NST in terms of oxidative stress could be indirect; by limiting the thermal dependency of animals upon the shivering process during prolonged cold exposure. Indeed, muscular shivering thermogenesis relies on muscle contractile activity, which itself relies on strong ATP production to fuel this activity. Contractile activity was previously reported to be positively related to ROS production and to a transient decrease in thiols content, followed by increased levels of various antioxidant enzymes (McArdle et al., 2001). Our results show that both WT and UCP1-deficient mice exhibited an increased proportion of oxidized glutathione after cold exposure, with a significantly higher level in UCP1-KO mice. Given that muscle GR activity reached the same level in both groups in cold conditions (see supplementary material Fig. S1) and that total glutathione did not significantly differ between groups, it implies that ROS production of muscle mitochondria might have been increased in the cold. Nevertheless, a direct measurement of muscle ROS production is required to ascertain this hypothesis. This potential increase in ROS production seems to have a different final impact (i.e. oxidative damage) depending on mice genotype. The slight rise of ROS production in WT mice, which could be attributed to a low or transient shivering activity or to a switch in pro-oxidant metabolic substrate (i.e. lipid mobilization) (St-Pierre et al., 2002), had no impact on protein carbonyl levels. On the contrary, UCP1-KO mice exposed to 12°C showed a larger oxidative imbalance and higher protein carbonyl content in skeletal muscle. Interestingly, recent work has demonstrated a rise in muscular mitochondrial ROS production for UCP1-KO mice acclimated to 5°C but not for WT mice, which supports our results (Oelkrug, 2013). Furthermore, the idea that NST can indirectly protect the muscle from an overloading ROS production is in agreement with previous studies reporting that muscle antioxidant enzyme activities decreased over time in WT mice exposed to cold (Petrovic et al., 2008) and that lifelong exposure to cold caused no significant muscle oxidative damage in wild derived rodents (Selman et al., 2008). Note also that it has recently been demonstrated that physical activity can induce the production of irisin by the muscle, and that this hormone stimulates UCP1 expression and a brown-fat-like development of white adipose cells (Boström et al., 2012). Hence, such a system may act as a negative feedback to mitigate the deleterious impact of prolonged muscle shivering, such as oxidative stress (present study) or defective calcium handling (Aydin et al., 2008).

Insights on mitochondrial functioning have shown that oxygen consumption is principally coupled with energy conversion as either ATP or heat, depending on whether the ATP synthase or the mitochondrial UCP1 is driving respiration. There is, however, growing evidence that these two processes might lead to differing oxidative costs (Brand, 2000). In accordance with one common expectation of the ‘free radical theory of ageing’, our results show that the high metabolism of UCP1-KO mice acclimated to cold, which is coupled to high ATP-dependent muscular shivering thermogenesis, is associated with increased levels of oxidative stress/damage in the muscles and in the blood. Alternatively, and in agreement with expectations of the ‘uncoupling to survive hypothesis’, we found that the cold-induced activation of UCP1 in the BAT (i.e. NST) allowed WT mice to increase their metabolism to generate heat while preventing them from oxidative damage. Therefore, we suggest that determining the accurate nature of the mitochondrial mechanisms implicated not only in the control of metabolism in a given environmental condition (present study), but also in the determination of life history trajectories (Salin et al., 2012a; Salin et al., 2012b), is an important milestone in our understanding of the determinants of longevity.

The study complied with the ‘Principles of Animal Care’ publication no. 86-23, revised 1985 of the National Institutes of Health, and with current legislation (L87-848) on animal experimentation in France. The experiment started with 40 non-reproductive male and female mice C57 black 6 from our animal husbandry unit (temperature=26±1°C). Half of the animals were wild type (WT) mice and the other half were UCP1-knockout mice (UCP1-KO). The founder mice (C57BL/6 J) UCP1-KO for establishing our colony were originally provided by the CNRS (UPR-9078) and were backcrossed and genotyped according to The Jackson Laboratory protocol. During 5 weeks, 10 mice per genotype were maintained at 26°C (groups WT26 and KO26) and 10 mice per genotype were exposed to 12°C (groups WT12 and KO12) for 4 weeks, after 1 week of progressive cooling (2°C per day). The cold exposure was chosen to be moderate (12°C for 4–5 weeks) for two main reasons: (1) to avoid premature death of UCP1-KO mice since their longevity is markedly reduced when exposed to 4°C (Golozoubova et al., 2001) and (2) to avoid differences between WT and KO mice in terms of metabolic rate since it has been shown that below 12°C UCP1-KO mice might exhibit higher metabolic rates (Ukropec et al., 2006). We used an equal number of male and female mice in each experimental group (five males/five females) and the animals did not differ between groups in terms of mass [general linear model (GLM), P=0.97] and age (GLM, P=0.49) at the beginning of the experiment. Animals were maintained on a 12 h:12 h light:dark light cycle, and food (SAFE A03) and water were provided ad libitum.

At the end of the experimental period, animals were culled (between 13:00 h and 16:00 h to restrict circadian bias in oxidative stress parameters) by cervical dislocation followed by decapitation in order to collect blood in heparinised micro tubes, as well as to collect the BAT and skeletal muscles (i.e. a mix of thigh and abdominal muscles). Immediately after collection, blood samples were centrifuged (3000 g for 10 min) to separate plasma from cells, and tissue samples were snap-frozen in liquid nitrogen. Samples were subsequently stored at −80°C until laboratory analysis.

Oxygen consumption (, expressed in ml O₂ consumed per min) was determined twice for eight animals of the WT12 and KO12 groups. The first measurement was taken before the experimental period, close to thermoneutrality (26°C), and the second was taken 1 week before the end of the experiment during the moderate (12°C) cold exposure. We recorded O₂ consumption (open-circuit indirect calorimetry system, Sable Systems, Las Vegas, NV, USA) for 5 h after one night of acclimation (without food but with water ad libitum). We used the average of these 5 h to obtain the mean .

The antioxidant capacity and the concentration of reactive oxygen metabolites (ROMs) were measured using the OXY-adsorbent (5 μl of 1:100 diluted plasma) and d-ROMs (5 μl of plasma; Diacron International, Grosseto, Italy) tests following the manufacturer's protocol. The OXY-adsorbent test quantifies the ability of the plasma antioxidant components to buffer massive oxidation through hypochlorous acid, while the d-ROMs test measures mostly hydroperoxides as a marker of global early oxidative damage [for a review of the literature of previous experiments using those two markers of oxidative stress, see Stier et al. (Stier et al., 2012)]. Antioxidant barrier is expressed as mmol l⁻¹ of HClO neutralised, and d-ROMs is expressed as mg of H₂O₂ equiv dl⁻¹. Mean ± s.e.m. intra-individual coefficient of variation based on duplicates was 2.23±0.30% for the OXY test and 1.90±0.26% for the d-ROMs test. Inter-plate coefficient of variation based on a standard sample repeated over plates was 4.16% for the OXY test and 2.75% for the d-ROMs test.

Glutathione content and the proportion of oxidized glutathione in BAT and muscle was determined using the DetectX® glutathione fluorescent detection kit (Arbor Assays, Ann Arbor, MI, USA), following manufacturer's instructions. Glutathione plays a key role in many biological processes, including the protection of cells against oxidation. Glutathione is used as a reductant by the enzyme glutathione peroxidase to scavenge deleterious hydrogen peroxide. The oxidized form of glutathione (GSSG) can be restored into glutathione by the action of the enzyme glutathione reductase. We evaluated the total glutathione content, as an indicator of antioxidant protection, and the ratio GSSG/total glutathione (which represents the proportion of oxidized glutathione), as an indicator of oxidative challenge (i.e. the pro-oxidant power buffered by the glutathione system). Values are expressed, respectively, as nmol total glutathione mg⁻¹ protein and as a ratio of oxidized glutathione/total glutathione [0 meaning that all glutathione is free glutathione, and 1 meaning that all glutathione is oxidized (GSSG)]. Mean ± s.e.m. intra-individual coefficient of variation based on duplicates was 4.13±0.42%.

To assess oxidative damage of protein in BAT and skeletal muscle, we determined protein carbonylation using the Oxiselect™ protein carbonyl spectrophotometric assay kit (Cell Biolabs Inc., San Diego, CA, USA) following manufacturer's instructions. This method allows the quantification of carbonyl content, which is a common form of ROS-induced protein oxidation. All samples were measured on the same plate. Values are expressed as nmol protein carbonyl mg⁻¹ protein. Mean ± s.e.m. intra-individual coefficient variation based on duplicates was 5.13±0.78%. Total protein content of tissue homogenates was determined in duplicate using a Pierce™ BCA protein assay (Thermo Scientific, Waltham, MA, USA).

We investigated genotype and temperature effect on metabolic rate () by running a repeated-measures GLM. We used individual as subject, temperature as within-subject factor, genotype and the interaction between genotype and temperature as fixed factors, and mass as a covariate.

We investigated the effects of genotype (WT versus KO), temperature (26°C versus 12°C), and the interaction between genotype and temperature on oxidative stress parameters with GLMs, after testing residuals of each model for normality and homoscedasticity. When a significant interaction between genotype and temperature was revealed, we ran a post hoc analysis to determine statistical differences between our four experimental groups.

Age and sex were initially included in statistical models but were not significant; they were removed in order to clarify statistical models. Repeated-measures GLMs were fitted with a normal error distribution (SPSS 18.0). Analyses were two-tailed tests and P-values ≤0.05. Means are quoted ±s.e.m.

We are grateful to two anonymous reviewers for providing interesting and constructive comments on a previous draft of the paper.