|Home | About | Journals | Submit | Contact Us | Français|
Recently thioredoxin reductase 1 (TrxR1), encoded by Txnrd1, was suggested to modulate glucose and lipid metabolism in mice. Here we discovered that TrxR1 suppresses insulin responsiveness, anabolic metabolism and adipocyte differentiation. Immortalized mouse embryonic fibroblasts (MEFs) lacking Txnrd1 (Txnrd1−/−) displayed increased metabolic flux, glycogen storage, lipogenesis and adipogenesis. This phenotype coincided with upregulated PPARγ expression, promotion of mitotic clonal expansion and downregulation of p27 and p53. Enhanced Akt activation also contributed to augmented adipogenesis and insulin sensitivity. Knockdown of TXNRD1 transcripts accelerated adipocyte differentiation also in human primary preadipocytes. Furthermore, TXNRD1 transcript levels in subcutaneous adipose tissue from 56 women were inversely associated with insulin sensitivity in vivo and lipogenesis in their isolated adipocytes. These results suggest that TrxR1 suppresses anabolic metabolism and adipogenesis by inhibition of intracellular signaling pathways downstream of insulin stimulation.
Adipose tissue plays a major role in energy metabolism, and aberrations in its development or function can lead to an array of metabolic alterations and diseases. Up to 90% of the volume of adipose tissue is comprised of adipocytes generated through adipogenesis from specific precursor cells and in adult humans, white adipose tissue is in a dynamic state where approximately 10% of the adipocyte pool is turned over annually1. The process of adipogenesis requires an orchestrated multistep process of sequentially concerted signaling events, converging on activation of key transcription factors required for adipocyte formation. The most critical transcription factors include members of the CCAAT-enhancer-binding protein (C/EBP) and peroxisome proliferator-activated receptor (PPAR) families2,3. Early signaling events mediated by C/EBPβ and C/EBPα contribute to initiate adipogenesis by inducing PPARγ, among others. In addition, they also contribute to maintenance of the adipose phenotype3. Activation of the nuclear receptor PPARγ is both necessary and sufficient for differentiation of adipocytes4 and pro-adipogenic or anti-adipogenic factors either induce or repress PPARγ, respectively3. Adipocyte differentiation is also closely linked to insulin signaling3. Attenuating insulin signaling, for instance by loss of insulin-receptor substrate (IRS) proteins, inhibition of phosphatidylinositol-3 kinase (PI3K) or depletion of AKT/protein kinase B (PKB), leads to suppression of adipocyte differentiation5,6,7.
The mammalian selenoprotein thioredoxin reductase 1 (TrxR1), encoded in mice by Txnrd1 and in human by TXNRD1, plays a key role in antioxidant defense, redox regulation, cell proliferation and cell signaling events, by catalyzing reduction of the active site disulfide of thioredoxin 1 (Trx1) and several other substrates in cells8. Germline deletion of Txnrd1 in mice causes early embryonic death9,10. However, hepatocyte-specific conditional deletion of Txnrd1 is not lethal, but results in pronounced alterations of glycogen and lipid storage in the liver11,12. Although somewhat conflicting data have been published, some observations indicate that TrxR1 can influence lipid turnover. Hepatocyte-specific disruption of Txnrd1 was found in one study to cause a metabolic switch in which hepatic lipogenesis seemed to be repressed and glycogen storage greatly increased11, while another study reported mild to severe hepatic accumulation of lipids12. In the study by Iverson and colleagues, lipid content was assessed using transmission electron microscopy (TEM) and appeared repressed in periportal hepatocytes because of high glycogen accumulation11. In the Carlson study, lipids were assessed using morphological assessments of hepatocyte vacuoles that were judged to resemble lipid vesicles12. Thereby neither of these studies validated fat accumulation by direct measurements of lipid content. Thus, while both studies suggested a role of TrxR1 in regulation of glucose and/or lipid metabolism, effects of TrxR1 on lipid metabolism clearly remain to be defined.
Here, we first examined the impact of TrxR1 on glucose and lipid metabolism using well-defined cell culture models optimally suited for studies of molecular mechanisms. Because primary MEFs can be differentiated into osteocytes, chondrocytes or adipocytes depending upon choice of hormonal stimuli13, we studied the propensity of Txnrd1−/− MEFs to undergo adipogenesis and compared this to parental cells expressing the intact Txnrd1 gene. We also determined adipocyte differentiation of primary human preadipocytes transfected with TXNRD1-targeting or non silencing siRNA oligonucleotides, and analyzed possible correlations of TXNRD1 transcript levels to clinical parameters in a cohort of obese and non-obese women. Our results collectively suggest that TrxR1 exerts a hitherto unknown but potent role in regulation of insulin responsiveness and adipogenesis.
The Txnrd1−/− MEFs used herein were previously characterized in detail and confirmed to completely lack TrxR1 expression while, in contrast, the parental Txnrd1fl/fl MEFs have a TrxR activity of ≈25nmol/min/mg protein14,15. Both cell types are immortalized batches of MEF, where the Txrnd1−/− genotype was obtained by ex vivo treatment of the Txnrd1fl/fl cells with Tat-Cre14,15. Here we observed that glucose uptake and glycolytic flux were not affected by deletion of TrxR1 (Fig. 1A,B). However, basal mitochondrial respiration rates in the presence of glucose, but not maximal respiration or unproductive (non-ATP coupled) oxygen consumption, were significantly increased in Txnrd1−/− MEFs (Fig. 1C), indicating higher metabolic flux without increased mitochondrial capacity or any apparent mitochondrial dysfunction. The increase in basal mitochondrial respiration was not due to an expansion in mitochondrial mass, which was comparable between Txnrd1fl/fl and Txnrd1−/− cells (Fig. 1D,E). Strikingly, increased mitochondrial respiration in Txnrd1−/− cells was furthermore not correlated with higher steady-state ATP levels (Fig. 1F). This led us to hypothesize that the elevated metabolic flux in absence of TrxR1 was geared towards anabolic biosynthetic processes. Indeed, we observed that the Txnrd1−/− cells contained significantly higher levels of glycogen (Fig. 1G) as well as triglycerides (Fig. 1H) compared to Txnrd1fl/fl MEFs.
The significantly higher content of triglycerides in Txnrd1−/− MEFs showed that Txnrd1 deletion facilitated fat accumulation. Remarkably, cultures of Txnrd1−/− MEFs, unlike Txnrd1fl/fl, displayed a readily observable proportion of cells with adipocyte-like morphology containing small lipid droplets as confirmed by Oil Red-O staining (Fig. 2A,B). We therefore asked whether the enhanced ability of Txnrd1−/− MEFs to store fat could correlate with an increased propensity for differentiation into adipocytes. To test this hypothesis, we attempted to induce adipogenic differentiation by hormonal induction using a well-established combination of pro-adipogenic factors (DMI: dexamethasone, methylisobutylxanthine and insulin, together with rosiglitazone). Indeed, Txnrd1−/− MEFs displayed evident signs of adipocyte differentiation, which were not observed in the parental cells (Fig. 2A–C).
Because PPARγ and C/EBPα are crucial transcriptional activators in adipogenesis, we next analyzed their levels in the Txnrd1−/− and Txnrd1fl/fl MEFs. We found that both PPARγ and C/EBPα were readily detected in Txnrd1−/− cells and further elevated by adipogenic induction, whereas their expression levels were much lower in the parental Txnrd1fl/fl cells (Fig. 2D). We also examined the expression of adipocyte fatty-acid-binding protein 4 (FABP4/aP2), which is a widely used adipogenic marker16. Pronounced expression of FABP4/aP2 was found in hormonally induced Txnrd1−/− MEFs, but neither in the untreated Txnrd1−/− cells nor in the parental Txnrd1fl/fl MEFs irrespective of hormonal treatment (Fig. 2D). Uncoupling protein 1 (UCP1) was also spontaneously expressed in the Txnrd1−/− cells, and upregulated in both knockouts and parental cells after hormonal induction but to a higher extent in the Txnrd1−/− MEFs (Fig. 2D). These results show that Txnrd1−/− MEFs has a more efficient adipogenic program compared to immortalized wild type Txnrd1fl/fl MEFs.
As we did not observe adipocyte differentiation of the parental Txnrd1fl/fl MEFs, possibly due to the immortalization process that is well known to deprive MEFs of their potential to undergo adipogenesis3, we also analyzed freshly isolated primary MEFs and compared them with the Txnrd1−/− MEFs. This demonstrated that, unlike the immortalized Txnrd1fl/fl MEFs, primary MEFs were indeed able to differentiate into adipocytes, but only after a very long period of hormonal induction (23 days). Adipocyte differentiation of the Txnrd1−/− MEFs was thus clearly enhanced compared also to primary MEFs, as illustrated by their stronger Oil Red-O staining and higher levels of FABP4/aP2 expression (Fig. 2E,F). Enzymatic activity assays on cell lysates harvested at different time points showed that cellular TrxR activity levels correlated inversely with the adipocyte differentiation capability of these cell types, with Txnrd1fl/fl MEFs having 2-fold higher basal TrxR activity than primary MEFs, while the knockout cells only retained background activity, most likely representing mitochondrial thioredoxin reductase (TrxR2) activity15 (Fig. 2G).
Post-confluent and growth-arrested pre-adipocytes reenter the cell cycle to undergo two sequential rounds of mitosis following hormonal induction of differentiation, which is referred to as mitotic clonal expansion (MCE)17, a necessary step for adipogenesis in MEFs18. Here we found that the cell number in Txnrd1−/− MEF cultures, in the absence of hormonal induction, still increased after reaching confluence to almost 2-fold after 96h (Fig. 3A, left panel). Cell number was further increased upon hormonal induction of adipocyte differentiation (Fig. 3A, right panel). In contrast, there was no cell proliferation beyond confluency in Txnrd1fl/fl MEFs, irrespective of hormone treatment (Fig. 3A). In agreement with these cellular phenotypes, the MCE-related transcription factor CCAAT-enhancer-binding protein β (C/EBPβ)18 was highly induced in Txnrd1−/− MEFs following induction of differentiation, while this increase was much less pronounced in the parental cells (Fig. 3B). Interestingly, both the adipogenesis promoting (LAP, liver-enriched activator protein) and antiadipogenic (LIP, liver-enriched inhibitory protein) isoforms of C/EBPβ19 were increased in Txnrd1−/− MEFs, but at the early time points LAP was more increased than LIP (Fig. 3B). Strong upregulation of cyclin A at 24h after induction in the knockout cells, but not in the parental cells, was in agreement with the finding that only Txnrd1−/− MEFs traversed the G1-S checkpoint after having reached confluency (Fig. 3B). These findings suggest that TrxR1 negatively regulates adipogenesis by facilitating contact inhibition and suppression of MCE.
Because insulin signaling has strong modulatory effects on adipogenesis we next analyzed the phosphorylation status of key intracellular phosphorylation cascade proteins linked to insulin signaling. The serine/threonine kinase Akt (also known as protein kinase B, PKB) is involved in most of the metabolic actions of insulin20. Here, we observed a transiently enhanced phosphorylation of Akt at residues Thr308 and Ser473 within the first hour after hormonal induction of differentiation, which was more pronounced in Txnrd1−/− MEFs compared to the parental cells (Fig. 3C). Phosphorylation of cyclic AMP response element–binding protein (CREB), which is linked to Akt activation21 and important for adipogenesis signaling22, revealed no major difference between the two MEFs except for a slightly more sustained expression in Txnrd1fl/fl cells at the 120min time point (Fig. 3C). However, activating transcription factor 1 (ATF1), which is closely related to CREB and shares similar functional motifs23, was more phosphorylated in Txnrd1−/− MEFs after hormonal induction of differentiation (Fig. 3C). Another effector of insulin action involved in adipogenesis is mTOR124, the activity of which is illustrated by the levels of phosphorylation of its downstream target ribosomal protein S625. Intriguingly, we found that the baseline phosphorylation level of S6 was lower in Txnrd1−/− MEFs compared to the parental cells, but more responsive with strongly increased phosphorylation upon induction of differentiation (Fig. 3C). These results collectively demonstrate clearly enhanced intracellular insulin responsiveness after deletion of the Txnrd1 gene.
Previous studies have shown that the Trx system helps to reduce and thus activate oxidized protein-tyrosine phosphatase 1B (PTP1B) in relation to PDGF signaling26. Because PTP1B is an important negative regulator also of insulin signaling, as it dephosphorylates key tyrosine residues of the insulin receptor (IR)27, we next assessed IR phosphorylation status. We observed no major difference in phosphorylation of tyrosine residues on IR when comparing parental and Txnrd1−/− MEFs after rosiglitazone treatment, with similarly increased phosphorylation of IR in both cell types (Fig. 3D). This suggests that the difference in insulin responsiveness between Txnrd1−/− and Txnrd1fl/fl MEFs should be explained by signaling events downstream of IR activation, which was analyzed next.
Insulin signaling requires phosphatidylinositol (PI) 3-kinase (PI3K)-mediated phosphatidylinositol (3,4,5)-trisphosphate (PIP3) generation, which in turn activates Akt/PKB and this process is negatively regulated by the phosphatase and tensin homologue (PTEN) protein20. Here we found that PTEN protein levels were lower in Txnrd1−/− cells compared to the parental Txnrd1fl/fl MEFs (Fig. 4A). Many factors are known to regulate PTEN expression, including p5328 that upregulates PTEN at both the mRNA and protein levels29,30. In mammalian cells, TrxR inhibition was shown to impair p53 conformation and function31, which prompted us to analyze p53 levels in our cell models. Interestingly, p53 was hardly detectable in Txnrd1−/− MEFs, regardless of stages of cell growth or upon doxorubicin treatment, which normally increases p53 expression. In contrast, p53 levels were clearly responding in Txnrd1fl/fl MEFs with a higher expression during proliferation or upon doxorubicin treatment (Fig. 4B,C). It is thus possible that p53 can contribute to the higher PTEN levels in Txnrd1fl/fl cells compared to Txnrd1−/−, but it is known that PTEN can also be regulated by several different mechanisms.
It was reported that Trx1 can bind to and inhibit PTEN in a redox dependent manner, which may result in activated Akt32, but, conversely, Trx1 can also reduce oxidized PTEN to increase its activity33,34. We therefore investigated the levels of Trx1 as well as its redox status. We first confirmed that Trx1 protein levels are upregulated in Txnrd1−/− MEFs upon hormonal treatment (Fig. 4D), as reported previously for the regularly propagated cells15. In our earlier study, we showed that Trx1 was mostly present in its reduced form in Txnrd1−/− cells, unless the cells were challenged with exaggerated oxidative stress15. Here we found that Trx1 was maintained mostly in its reduced form in Txnrd1−/− MEFs also after hormonal induction of adipocyte differentiation (Fig. 4E). Co-immunoprecipitation studies furthermore showed that a physical interaction between Trx1 and PTEN could be detected in Txnrd1−/− cells, but not in the parental MEFs (Fig. 4F,G). Taken together, these results would be compatible with Trx1-mediated binding to and inhibition of PTEN in the Txnrd1−/− cells, which may further contribute to their sensitization to insulin signaling. To examine this possibility, we transiently overexpressed wildtype (wt) PTEN, or its inactive C124S mutant derivative, in the Txnrd1−/− MEFs. This showed that overexpression of wt PTEN counteracted the adipogenesis of Txnrd1−/− cells, as illustrated by lower levels of the adipocyte marker FABP4/aP2 and less Oil Red-O stained lipid accumulation, with no major effects on PPARγ levels (Fig. 4H). Collectively, these findings suggest that attenuated PTEN activity due to lower PTEN levels and inhibitory binding of Trx1 should be a part of the molecular mechanisms explaining the augmented adipogenesis observed in Txnrd1−/− MEFs.
Since physiological increases in H2O2 levels have been proposed as important triggers of adipocyte differentiation35 and because we found earlier that Txnrd1−/− MEFs present higher levels of H2O215, we next investigated if supplementation with antioxidants could blunt the increased adipogenesis process in these cells. Indeed, PEG-catalase and α-tocopherol attenuated adipocyte formation and lipid accumulation. Strikingly, N-acetyl cysteine (NAC) supplementation completely abolished adipogenesis in Txnrd1−/− MEFs (Fig. 5A,B). None of these treatments affected the expression levels of PPARγ (Fig. 5C). However, in agreement with lower adipogenesis capability (Fig. 5A), the treatments inhibited FABP4/aP2 expression, with NAC totally abrogating its expression (Fig. 5C). We furthermore found that NAC did not increase, but rather decreased, the total glutathione (GSH+GSSG) levels in Txnrd1−/− MEFs, although these were still higher than in the parental Txnrd1fl/fl MEFs (Fig. 5D). Thus, the blocking effects of NAC were not necessarily due its function as a cysteine and GSH precursor, and could involve several of the other reported effects of NAC supplementation36. We further found that NAC, but not PEG-catalase, almost completely abolished the capacity for MCE of Txnrd1−/− MEFs (Fig. 5E), which should be part of the explanation how NAC completely inhibited adipogenesis in these cells. However, neither NAC nor PEG-catalase affected C/EBPβ expression after hormonal induction (Fig. 5F). Upregulation of cyclin A in Txnrd1−/− MEFs was however attenuated by NAC treatment, which agrees with its blocking effect on MCE (Fig. 5F).
We next investigated NAC effects on transcription levels of cyclin-dependent kinase inhibitors and cyclins. This revealed that p27, an inhibitor of cell cycle progression37, was downregulated during initiation of adipocyte differentiation in Txnrd1−/− MEFs compared to parental cells, but that NAC treatment attenuated the fluctuations in its transcript levels at later time points (Fig. 5G). Conversely, the transcript levels of cyclin E were higher in Txnrd1−/− MEFs and increased to a larger extent on day 3 compared to Txnrd1fl/fl MEFs, with NAC partially attenuating this difference (Fig. 5G). To further validate these findings, we analyzed the protein levels of p27 and cyclin E at different time points between day 2 and day 4 after hormonal induction of adipogenesis (Fig. 5H). Reflecting the effects on mRNA levels, the p27 protein levels were much lower in Txnrd1−/− compared to Txnrd1fl/fl MEFs and further decreased on day 3, whereas NAC treatment increased the p27 protein levels in these cells (Fig. 5H). Cyclin E was also higher in the knockout MEFs compared to parental cells, with NAC slightly lowering its expression towards the later time points (Fig. 5H). Akt activation has been reported to depress, both directly and indirectly, the expression levels of p2738,39. Here we indeed found that Akt was more activated in the knockout cells, as indicated by phosphorylation of its Ser473 residue, while NAC prevented this activation (Fig. 5H).
We next found that Txnrd1−/− MEFs displayed as much as 30-fold higher basal PPARγ expression levels than the parental cells, which were further elevated upon adipocyte differentiation. NAC treatment somewhat inhibited upregulation of PPARγ levels in the initial days of the differentiation process, but not nearly down to the levels found in the parental cells (Fig. 6A). Not only PPARγ expression levels but also extent of activation can thereby contribute to the observed phenotypes, which we studied next.
Nitration modified unsaturated fatty acids are endogenous PPARγ ligands that can be induced upon oxidative stress40,41. We thus next assessed lipid peroxidation levels in our cell models using a Bodipy (581/591) reagent. Txnrd1−/− MEFs had six times higher levels of lipid peroxides compared to Txnrd1fl/fl MEFs, which were to some extent dampened by treatment with the lipid soluble antioxidant α-tocopherol. Treatment with NAC, in contrast, failed to reduce lipid peroxidation in the knockout MEFs but instead increased the signal (Fig. 6B). These findings suggested that the very high expression of PPARγ together with higher lipid peroxide levels could provide a mechanism for increased adipogenic potential of Txnrd1−/− MEFs, but the potential involvement of nitrated lipids and their possible prevention by NAC was still unclear.
To investigate direct effects of nitro-modified unsaturated fatty acids, we attempted to stimulate differentiation of the Txnrd1−/− MEFs using either 9-Nitrooleate (AONO2) or 10-Nitrolinoleate (LNO2) instead of using the standard PPARγ ligand rosiglitazone in the hormonal induction cocktail. Interestingly, treatment of Txnrd1−/− MEFs with dexamethasone, methylisobutylxanthine and insulin alone (DMI) increased expression of PPARγ and FABP4/aP2, while addition of AONO2, but not of LNO2, further promoted adipocyte formation as illustrated by increased FABP4/aP2 expression (Fig. 6C). Notably, neither of these two compounds upregulated the PPARγ expression levels more than DMI treatment alone (Fig. 6C). These findings agree with the notion that AONO2 directly activates PPARγ as a ligand40. Because nitro-modified unsaturated fatty acids are highly electrophilic and can interact directly with free thiols42 we asked whether NAC, having a readily accessible thiol group, might directly scavenge these electrophilic metabolites. If so, that could potentially block adipocyte differentiation as a result of inhibited PPARγ activation. Indeed, both AONO2 and LNO2, but not rosiglitazone, were found to readily react with the free thiol group of NAC at roughly stoichiometric amounts upon 30min incubation in vitro (Fig. 6D). If such direct scavenging would occur also in of Txnrd1−/− MEFs, this should likely contribute to the antiadipogenic effect of NAC treatment.
To probe if TrxR1 expression counteracts adipogenesis also in primary human cells, we isolated stroma vascular cells from abdominal subcutaneous adipose tissue that were subsequently cultured and induced to differentiate in vitro. Following hormonal induction (DMI+Rosi), we found that cells treated with siRNA-mediated TXNRD1 knockdown significantly increased their adipogenesis, as demonstrated by a stronger Oil Red-O staining, compared to cells that had been transfected with non-silencing siRNA. This was seen with cells from two different donors and either with or without selenium supplementation to the growth medium (Fig. 7A–C).
To assess whether TXNRD1 expression in adipose tissue could be related to clinical and/or adipocyte-specific aspects of insulin sensitivity, we analyzed data from a global transcriptional profiling study in 56 women (30 obese and 26 non-obese)43. We found that TXNRD1 levels correlated with body mass index (BMI, r=0.56, p<0.0001). Thus, to avoid a bias for BMI as a confounding factor, multiple regression analyses were performed. This revealed that TXNRD1 transcript levels in adipose tissue were inversely associated with in vivo measures of insulin responsiveness (determined with an intravenous insulin tolerance and plasma glucose reduction test), as well as adipocyte responsiveness to insulin in vitro (determined as insulin-stimulated lipogenesis at maximal effective concentration). Notably, these relationships were specific for insulin responsiveness as there were no BMI-independent correlations between TXNRD1 and other measures of lipid or adipocyte metabolism such as e.g. plasma triglyceride levels or adiponectin secretion (Table 1). These findings revealed that, similarly to our findings in the MEF cell models, TXNRD1 expression levels correlated with insulin responsiveness and lipogenesis also in this human cohort.
We recently found that MEFs lacking TrxR1 (Txnrd1−/−) undergo massive cell death when cultured at low density (<1000cells/cm2) in high-glucose medium, due to impaired self-sufficient growth and insufficient elimination of glucose-derived H2O215. However, when cultured at higher density (>8000cells/cm2) these cells grow well due to a release of catalase activity to the medium sufficient to detoxify their higher levels of metabolism-derived H2O215. Txnrd1−/− cells could thus be utilized herein as a well-defined cell culture model system for in-depth investigations concerning metabolic consequences of TrxR1 removal. Here we demonstrated that Txnrd1 deletion channels glucose usage towards increased basal mitochondrial respiration and anabolic pathways, resulting in increased glycogen and lipid biosynthesis. The differences in basal mitochondrial respiration but not in spare respiratory capacity could be due to effects on respiratory complex assembly or perhaps redox-derived post-translational modifications of respiratory subunits. Interestingly, recent evidence has shown how oxidative stress can trigger signaling pathways that modulate respiratory rates44. Our results clearly warrant further research to delineate the role of TrxR1 in regulating mitochondrial respiration. We also showed that deletion or knockdown of TrxR1 strongly improves the capacity of immortalized MEFs or human pluripotent primary cells to undergo adipogenesis. The highly significant inverse association between TXNRD1 transcript levels in human adipose tissue and insulin responsiveness in vivo and in vitro suggests that our findings have relevance for human physiology. From our more detailed analyses of Txnrd1−/− MEFs, insights with regards to the molecular mechanisms involved could be gained. These mechanisms are graphically summarized in Fig. 8 and discussed further as follows.
Insulin-triggered intracellular signaling cascades are known to play essential roles in adipocyte differentiation3. Our findings of stronger activation of Akt and CREB/ATF1 in the Txnrd1−/− MEFs correlate well with a more pronounced adipocyte differentiation phenotype being accentuated by exaggerated insulin signaling. As the phosphotyrosine phosphatase PTP1B is known to be an important negative regulator of insulin signaling by catalyzing dephosphorylation of IR27, and because PTP1B is reduced and activated by the TrxR1-dependent Trx system26, we were initially surprised not to detect any significant differences in IR phosphorylation status between Txnrd1−/− and Txnrd1fl/fl MEFs. One explanation might be that in the previous study where the Trx system was found to act in PTP1B activation, Txnrd1−/− MEFs were starved for 24h using 1% FBS and thereupon induced with platelet-derived growth factor (PDGF)26. Starvation is well known to induce oxidative stress and may thus have exaggerated PTP1B oxidation in the absence of TrxR1. Also, here we studied IR status and not the PDGF receptor and the exact ligand-receptor mediated phosphorylation cascades where the Trx system-catalyzed activation of PTP1B is important need to be further studied. Our findings in the present study however strongly suggest that the increased capacity for adipocyte differentiation of MEFs upon Txnrd1 deletion should be explained by accumulated effects of several altered signaling events downstream of IR that lead to increased insulin responsiveness. Such effects are likely to contribute to the inverse correlation between TXNRD1 transcription and clinical parameters of insulin action as found in our human cohort of obese and non-obese women, even though detailed mechanistic studies must be performed also with human cells and tissues before such assumptions can be fully confirmed.
Our finding that the expression level of the negative Akt regulator PTEN was lower in the TrxR1 deficient MEFs should be important for the observed phenotype, with the PI3K/Akt axis being crucial for insulin triggered adipogenesis45. Further studies are needed to validate whether the low p53 levels in Txnrd1−/− cells are an important factor explaining their propensity to differentiate. That would fit with earlier observations that p53 can be an efficient suppressor of white adipocyte differentiation46.
The interplay between the Trx system and regulation of PTEN activity is likely to be complex. PTEN is a redox sensitive phosphatase, which becomes inactivated upon exposure to H2O2 and formation of a disulfide bond between Cys71 and Cys12433,34. Trx1 has been shown to reduce and reactivate this type of oxidized PTEN33,34. However, another study found that the reduced form of Trx1 directly binds to PTEN to inhibit its activity32. We showed previously, and also here, that the Trx1 protein becomes upregulated in Txnrd1−/− MEFs, which can at least in part be due to Nrf2 activation15. Importantly, here we show that Trx1 can be maintained in its reduced form also in the absence of TrxR1 and regardless of the state of adipocyte differentiation. Since GSH and Grx can constitute a backup system for Trx1 reduction when TrxR1 becomes impaired14,47,48, especially with the GSH-dependent enzyme systems being upregulated in the Txnrd1−/− MEFs15, this results in the apparent paradox that the levels of reduced Trx1 can become higher than normal in TrxR1 knockout cells. As our immunoprecipitation experiments suggested a stronger interaction between Trx1 and PTEN in Txnrd1−/− cells, this would be compatible with the notion of Trx1 binding to and serving as an inhibitor of PTEN32. The attenuated adipocyte differentiation seen upon overexpressed PTEN in the Txnrd1−/− MEFs further supports this model.
Primary MEFs are normally able to differentiate in vitro upon induction using treatment with pro-adipogenic factors3,49, whereas most immortalized MEFs have lost the capacity to differentiate3,49. Here we showed that the basal levels of TrxR activity in the parental immortalized Txnrd1fl/fl MEFs are two-fold higher than in primary MEFs. Upon long time treatment with proadipogenic factors, primary MEFs could differentiate into the adipocyte lineage, but the immortalised Txnrd1fl/fl MEFs could not, consistent with our findings that TrxR1 suppresses adipocyte differentiation. With Txnrd1 depletion restoring and enhancing the adipocyte differentiation capacity of immortalised MEFs, this became a vivid illustration of the importance of TrxR1 for regulation of adipogenesis in MEFs. Interestingly, the Txnrd1−/− MEFs seemed to differentiate into brown-like or beige adipocytes, as reflected by increased UCP1 expression3,50. Thus, TrxR1 may possibly also play specific roles in regulation of white vs beige adipocyte differentiation, which should clearly be studied further.
To understand the mechanisms underpinning the increased adipogenesis in Txnrd1−/− MEFs, it is important to note that MCE is considered essential for adipogenesis of MEFs and 3T3-L1 cells17,18. As the Txnrd1fl/fl MEFs were unable to undergo MCE, this could be a main reason for their inability to differentiate into adipocytes. It was in this context interesting that TrxR1 deficient MEFs spontaneously underwent MCE, with hormonal inducers further augmenting the process. As shown before, and also found here with regards to lipid peroxidation, depletion of Txnrd1 in MEFs leads to a heightened oxidative stress15. Increased production of H2O2 can be detrimental if excessive, but can facilitate adipocyte differentiation through enhancement of signal transduction and acceleration of MCE35,51,52. Increased H2O2 levels helping to trigger adipogenesis is compatible with our findings that supplementation with antioxidants inhibited adipocyte formation. It was particularly intriguing that NAC completely abrogated adipocyte differentiation, partially through blocking of the MCE. One possible part of that effect could involve increased p27 levels, as p27 is well known to be a crucial protein regulated by oxidative events and causing cell cycle arrest53,54. Downregulation of p27 is also tightly connected to Akt activation39,55, which agrees with our findings in the Txnrd1−/− MEFs. Activated Akt phosphorylates p27 at threonine 157, which localizes p27 to the cytosol and leads to proteasomal degradation39,56. Akt also phosphorylates FOXO3a and inhibits its transactivation of p27 expression55. As NAC treatment blocked Akt activation in the knockout cells, it appears that oxidative stress can contribute to Akt activation in these cells, as shown earlier in other cell types57,58.
The importance of PPARγ in relation to our findings should be its ligand-mediated activation and pluripotent roles in adipogenesis59. Nitro-modified unsaturated fatty acids are endogenous PPARγ ligands acting at sub-micromolar concentration40, of which the production becomes facilitated by increased oxidative stress41. As we detected higher lipid peroxidation levels in the Txnrd1−/− MEFs compared to the Txnrd1fl/fl cells, this suggests that more nitro-modified unsaturated fatty acids are indeed produced in the knockout cells, which subsequently may activate PPARγ. Indeed, DMI treatment without inclusion of the PPARγ ligand rosiglitazone could still activate PPARγ and upregulate FABP4/aP2 expression in those cells, indicating that there must be an endogenously increased PPARγ ligand production in Txnrd1−/− cells. Since we further found that addition of AONO2 could promote adipocyte formation and that NAC could bind nitrated fatty acids directly in vitro, such scavenging may well have been another mechanism to explain the complete NAC blockage of the adipocyte differentiation that we observed.
In conclusion, our findings demonstrate that depletion of Txnrd1 promotes adipocyte differentiation of immortalized MEFs. This occurs through enhanced responsiveness to insulin signaling, with attenuated PTEN activity, loss of cell cycle control by activation of Akt, downregulation of p27, and activation of PPARγ. Our observations that TXNRD1 transcript levels inversely associate with clinical signs of human insulin resistance in vivo and in adipocytes in vitro warrant further studies addressing the potential role of TrxR1 as a therapeutic target in diseases affecting glucose and/or lipid metabolism. As suggested here, TrxR1 is as important modulator of cell fate and metabolism through suppression of insulin signaling and adipocyte differentiation.
The expression plasmid encoding wildtype (wt) PTEN (catalog number 10750) and mutant C124S PTEN (10744) with the control empty plasmid were all from Addgene (Cambridge, MA, USA) and have been described elsewhere60. 9-Nitrooleate (AONO2) and 10-Nitrolinoleate (LNO2) were purchased from Cayman Chemical (Ann Arbor, MI, USA). All chemicals were obtained from Sigma-Aldrich Chemicals (St. Louis, MO, USA) unless stated otherwise.
Studies in humans were approved by the regional ethical review board of Stockholm (Regionala Etikprövningsnämnden i Stockholm) and were conducted in full accordance with the Declaration of Helsinki. Informed written consent was obtained from all participants. Patient cells were obtained under ethical permit nr. 326/03 and array analyses were performed under ethical permit nr. 2011/1002-31/3. The mouse embryonic fibroblasts (MEFs) were isolated in Munich, Germany, in accordance with the German Animal Welfare Law and approved by the institutional committee on animal experimentation and the government of Upper Bavaria, Germany, under statements nr. VII 7/8730-1/6/99 and 211-2531.1-19/04.
The establishment of wildtype TrxR1 expressing parental MEFs having exon 15 of the Txnrd1 gene flanked by flox sites (Txnrd1fl/fl) and the full TrxR1 knockout MEFs derived from Txnrd1fl/fl cells after Cre treatment in vitro (hereafter referred to as Txnrd1−/−) were previously described15. The MEFs were kept at >10% confluence (split 1:8 or less for culture maintenance) in DMEM with 4.5g/l glucose content, supplemented with 4mM glutamine, 100U/ml penicillin, 100mg/ml streptomycin (BioWhittaker, Lonza, NJ, USA) and 10% (v/v) fetal bovine serum (PAA Laboratories, Piscataway, NJ, USA). No extra selenium source was added unless mentioned. Cells were grown in humidified air containing 5% CO2 at 37°C for all experiments. Primary MEFs were isolated from pregnant mice at day 13.5 post coitum. Single cells were obtained by shaking (100rpm) for 15minutes at 37°C with 0.05% trypsin/EDTA (Invitrogen), DNase (100U/ml, Roche) and glass pearls. The reaction was stopped by adding medium with 10% FCS. Cells were then cultured in DMEM medium mentioned above with 0.2% gelatin-coated flask. To induce differentiation, 2-day postconfluent MEFs (designated day 0) were cultured in DMEM containing 10% FBS, 1μM dexamethasone, 0.5mM methylisobutylxanthine, 5μg/ml insulin and 0.5μM rosiglitazone, unless stated otherwise. From day 2, medium containing 5μg/ml insulin and 0.5μM rosiglitazone was changed every other day until day 8 or as indicated.
Human adipose tissue was obtained from adult human subjects undergoing surgery for non-malignant conditions. Although both donors were anonymous to us, they had given informed written consent to donate tissue that would have otherwise been discarded. The study was approved by the regional board of ethics. Human adipose tissue-derived stromal vascular cells were isolated as previously described61 and cultured in 12-well format as previously described62 with one exception; differentiation was induced 48h instead of 24h after plating in order to downregulate TXNRD1 prior to initiation of adipogenesis. Gene expression of TXNRD1, which encodes TrxR1 in human, was knocked down using siRNA (sense 5′-(GCAAGACUCUCGAAAUUAU)dTdT-3′, antisense 5′-(AUAAUUUCGAGAGUCUUGC)dAdG-3′). Control cells were transfected with Allstars negative control (Qiagen, Hilden, Germany). In all cells, transfections were performed using HiPerFect transfection reagent (Qiagen, Hilden, Germany) with 20nM siRNA at two time points, 24h and 48h after plating. Differentiated cells were stained with Oil Red-O as described below 12–14 days after plating.
Cells were washed three times with PBS (Phosphate-buffered saline) and then fixed with 4% formaldehyde for 1h. Oil Red-O (0.3% in isopropanol) was diluted with water (3:2), filtered through a 0.22μm filter, and incubated with the fixed cells for 30min at room temperature. Cells were then washed with water, whereupon stained neutral fat droplets were visualized by light microscopy and photographed. Oil Red-O was thereafter extracted using isopropanol and absorbance was measured at 570nm for quantification.
The cells were seeded on 150mm dishes and cultured for 4 days to reach confluence. Triglyceride contents were then measured using a Triglyceride Quantification Kit (ab65336, Abcam, Cambridge, UK) according to the instruction manual.
The cells were seeded on 150mm dish and cultured for 4 days to reach confluence. Glycogen contents were measured using a Glycogen Assay Kit II (#ab169558; Abcam, Cambridge, MA, USA) according to the manufacturer’s guidelines.
The assessment of lactate production as an indication of glycolytic flux was performed as previously described with slight modifications63. Cells were seeded in M12 well plates, at a density of 50× 10E5 cells per well. The day after, the cells were changed to DMEM containing 2% FBS and 5mM glucose or 25mM glucose. Then 50ul of media were collected at 4h, 6h and 8h for analysis. At the end of the experiment, the cells were lysed (see protocol below for cell lysis) and the total amount of protein was determined for normalization. The amount of lactate present in the media was determined using the LACT (Lactate) reagent from Beckman Coulter (Brea, CA, USA), according to the manufacturer’s guidelines. An additional well per condition was ran in parallel with the glycolytic competitive inhibitor 2-deoxyglucose (30mM; Sigma-Aldrich, St. Louis, MO, USA). The amount of lactate generated in the presence of 2-deoxyglucose represents the non glycolytic-derived lactate, and was therefore substracted from the overall lactate production. The depicted results represent glycolysis-derived lactate production.
Oxygen consumption rate (OCR) in real time on MEFs was performed using an Extracellular Flux Analyzer X24-3 (Seahorse Bioscience, Billerica, MA, USA) as previously described63. Cells were seeded at a density of 50x 10E3cells/well. For the experiment the cells were changed to bicarbonate- and FBS-free DMEM medium (Sigma). The sequential addition of the mitochondrial ATP synthase inhibitor oligomycin (1uM; EMD Millipore, Billerica, MA, USA), the uncoupling ionophore carbonyl cyanide 4-trifluoromethoxyphenylhydrazone (FCCP; 0.5uM; Sigma-Aldrich) and the mitochondrial complex III inhibitor antimycin A (5uM; Sigma-Aldrich) allowed us to determine the fraction of respiration coupled to ATP production, the maximal respiratory capacity and the unproductive respiration not coupled to ATP production, also known as “leak”64.
Txnrd1fl/fl and Txnrd1−/− MEFs were seeded onto M24 wells and cultured for 1 day. The cells were then incubated with 500nM of MitoTracker Green (Life Technologies, Grand Island, NY, USA) for 15min at 37°C, washed twice with culture medium, and visualized using a Zeiss Axiovert 40 CFL fluorescent microscope (Zeiss, Overkochen, Germany), photographed with a coupled Zeiss Axiovert ICm1 camera. Fluorescence intensity was analyzed using the AxioVs40 v126.96.36.199 (Zeiss) and ImageJ v1.48 softwares. All samples were processed in parallel and images captured in the same session using the same imaging parameters. At least 500 cells per well and 3 wells per genotype were quantified for analysis. Data were expressed as Relative Fluorescent Units.
Cells were seeded at a density of 1.5×105cells/well in M6-well plates, and they were allowed to recover over night. Cells were starved for 5hours in DMEM without glucose, glutamine, pyruvate nor FBS for 5hours. The medium was then replaced by fresh medium containing 5ug/ml insulin, and the cells were incubated for 1hour in this medium. The medium was then replaced with 1ml of medium per well containing 5μg/ml insulin, 10μM 2-deoxyglycose (Sigma-Aldrich, St. Louis, MO, USA) and 1μCi (20nM) 3H-2-deoxyglucose (American Radiolabeled Chemicals, Inc., St. Louis, MO, USA). Cytochalasin B (Sigma-Aldrich) was used to inhibit insulin-independent glucose transport. Cells were allowed to uptake labeled 2-deoxyglucose for select periods of time (10, 20 and 30minutes), and then they were extensively washed with chilled PBS and lysed in ice-cold 0.4N NaOH. The lysates were mixed with Emulsifier Safe Scintillation liquid (Perkin-Elmer, Waltham, MA, USA) and 2-deoxyglucose uptake was quantified using a scintillation counter. Data were determined as counts per minute (CPM) and calculated in 2-deoxyglucose concentration using a known standard of 3H-2-deoxyglucose, and normalized by total amount of protein.
Antibodies against PPARγ (E-8), C/EBPα (14AA), C/EBPβ (C-19), insulin Rβ (C–19), p-Tyr (PY99), p27 (F-8), cyclin A (H-432), cyclin E (M-20), and GAPDH (FL-335) were from Santa Cruz Biotechnologies, Inc. (CA, USA). Antibody against FABP4/aP2 (ab66682) was obtained from Abcam Inc. (Cambridge, MA, USA). Antibodies against phospho-Akt (Thr308) (9275), Phospho-Akt (Ser473) (9271), Akt (9272), phospho-CREB (Ser133) (9191), phospho-S6 (Ser235/Ser236) (4858), S6 (2217), PTEN (9552) came from Cell Signaling Technology, Inc. (Danvers, Ma, USA). The antibodies against β-Actin (#A5441) and Voltage-Dependent Anion Channel-1 (VDAC1; #PC548) were purchased from Sigma-Aldrich (St. Louis, MO, USA) and EMD-Millipore, respectively. Anti-mouse Trx1 antibody serum was a kind gift from Dr. Gary Merrill (Oregon State University, Corvallis, OR, USA).
Cellular TrxR activity was measured using end-point insulin reduction assay. 4μg total cellular protein was incubated with 20μM recombinant human wt Trx, 297μM insulin, 1.3mM NADPH, 85mM Hepes buffer, pH 7.6, and 13mM EDTA for 80min at 37°C, in a total volume of 50μl. The reaction was then terminated by adding 200μl of 7.2M guanidine-HCl in 0.2M Tris-HCl, pH 8.0, containing 1mM DTNB. The extent of Trx-dependent formed thiols in the reduced insulin was then determined by measuring absorbance at 412nm (extinction coefficient 13,600M−1cm−1) and subtracting a background absorbance for each sample incubated and treated in the same manner containing all components except Trx.
Cells were harvested and lysed in RIPA buffer (50mM Tris-HCl, pH 7.4, 150mM NaCl, 1mM EDTA, 1mm EGTA, 1% Triton X-100, 1% sodium deoxycholic acid, 0.1% SDS, 0.5mM PMSF, 1.2mg/ml protease inhibitor cocktail (Roche, Basel, Switzerland), phosphatase inhibitor cocktail 3 (Sigma-Aldrich). The supernatants clarified after centrifugation (13300r.p.m., 15min) were used and total protein concentrations were determined with a Bradford reagent kit (Bio-Rad, Hercules, CA, USA). Protein lysates were subjected to SDS/PAGE and transferred onto nitrocellulose membrane using iBlot Dry Blotting System (Thermo Fisher Scientific, Waltham, MA, USA). For detection, the SuperSignal West Pico kit (Thermo Fisher Scientific, Waltham, MA, USA) was used according to the manufacturer’s instructions, with signals detected utilizing a Bio-Rad ChemiDoc XRS scanner and the Quantity One 4.6.7 software.
Cells were harvested and lysed in RIPA buffer as described above. First the primary rabbit antibody (2μg) was pre-incubated with 50μl of Dynabeads sheep anti rabbit IgG (Life Technologies, Grand Island, NY, USA) with gentle shaking for 3hours at 4°C. Total cell lysate (1 to 2mg protein) was then added into the mixture and incubated with gentle rocking overnight at 4°C. After several washes of the beads in PBS containing 0.1% BSA and 2mM EDTA, loading buffer was added to the samples and then they were boiled. Finally proteins in the supernatants were separated on 12% SDS–PAGE, transferred and probed. TrueBlot anti-rabbit IgG (Rockland Immunochemicals, Gilbertsville, PA, USA) was here used as secondary antibody.
Total RNA was isolated using the RNAeasy mini kit (Qiagen, Limburg, Netherlands) according to the manufacturer’s protocol. Total RNA (2μg) was reverse transcribed using Maxima First Strand cDNA synthesis kit (Thermo Fisher Scientific, Waltham, MA, USA) and random hexamer primers. QRT-PCR was performed using Maxima SYBR Green qPCR Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) on PikoReal 96 real time PCR system (Thermo Fisher Scientific, Waltham, MA, USA). The readouts for the mRNA levels of specific genes were normalized to 18S as housekeeping control. Specific primers were designed and synthesized (Thermo Fisher Scientific, Waltham, MA, USA) as follows: mouse p27: sense, 5′-CAGCTTGCCCGAGTTCTACT-3′; antisense, 5′-GAGTTTGCCTGAGACCCAAT-3′; mouse cyclin E: sense, 5′-CCCTTAAGTGGCGTCTAAGC-3′; antisense, 5′-TACTGAGGCATCAGCACCTC-3′; mouse PPARγ: sense, 5′-ATCTTAACTGCCGGATCCAC-3′, antisense, 5′-GATGGCATTGTGAGACATCC-3′; mouse 18S: sense, 5′-ACCGCAGCTAGGAATAATGGA-3′; antisense, 5′-GCCTCAGTTCCGAAAACCA-3′; human TrxR1: sense, 5′-ATATGGCAAGAAGGTGATGGTCC-3′; antisense, 5′-GGGCTTGTCCTAACAAAGCTG-3′; human LRP10: 5′-sense GATGGAGGCTGAGATTGTG-3′; antisense, 5′-GAGTCATATCCTGGCGTAAG-3′.
NAC (130μM) was incubated with different concentrations of AONO2, LNO2 or Rosiglitazone as indicated for 30min in PBS buffer in a 96 well plate. Then free thiols of NAC were quantified by adding 2mM DTNB and absorbance was measured at 412nm. The amount of free thiol was calculated using Beer-Lambert law and 14,150M−1cm−1 was used as the extinction coefficiency of TNB−65.
Txnrd1fl/fl and Txnrd1−/− MEFs were cultured to confluence, with the Txnrd1−/− MEFs incubated with or without 100μM α-tocopherol for two days. The cells were then stained with 5μM Bodipy (581/591) at 37°C for 30min. Cells were collected and fixed with 0.5% paraformaldehyde in PBS. Signals were analyzed using FACScan with the FITC channel and a total of 10000 cells of each sample were counted.
Redox immunoblotting was performed based on an electrophoretic mobility-shift assay47. Briefly, the cells were lysed with a sample solution (50mM Tris–HCl, 1mM EDTA, 8M urea, pH 8.3) containing 30mM iodoacetic acid (IAA). After incubation at 37°C for 30min, the proteins were precipitated and washed with ice-cold acetone/1M HCl (98/2v/v) three times. Then the precipitate was resuspended in the sample solution containing 3.5mM dithiothreitol (DTT) and incubated at 37°C for 30min. Subsequently, the proteins were alkylated with 10mM iodoacetamide (IAM) and separated by PAGE in buffer containing 8M urea and transferred onto nitrocellulose membrane using iBlot Dry Blotting System (Thermo Fisher Scientific, Waltham, MA, USA). Trx1 was detected using immunoblotting with antibodies against mouse Trx1. Form of Trx1 having more free thiols became more negatively charged by alkylation with IAA and thus migrated faster.
Cells were seeded at a density of 104cells/well in M96-well plates, and ATP levels were determined using the Cell Titer Glo reagent from Promega (Madison, WI, USA), according to the guidelines provided by the manufacturer, and expressed as relative luminescence units (RLU) and normalized by total amount of protein.
Data on TXNRD1 expression in subcutaneous abdominal adipose tissue was retrieved from a previously described cohort of 26 non-obese and 30 obese women analyzed by gene micro-array (Human Gene 1.0 ST Array, Affymetrix Inc., Santa Clara, CA)43. Global transcription profiling data have been described and are deposited at the National Center for Biotechnology Information Gene Expression Omnibus (http://ncbi.nim.nih.gov/geo) under the accession number GSE2540243.
Insulin sensitivity in vivo was determined by a short insulin tolerance test as previously described66. Briefly, plasma glucose levels after intravenous insulin injection were determined and plotted in a semilogaritmic graph. The rate constant (k) derived from this plot represented the intravenous insulin tolerance (KITT) and was expressed as the % fall in glucose/min between 4 and 16minutes. Plasma triglycerides (TGs) were determined by the hospital’s accredited clinical chemistry laboratory.
Lipogenesis in isolated fat cells was performed as described in detail previously67. In brief, isolated fat cells were incubated for 2hrs at 37°C in a buffer containing unlabelled glucose and 3H-glucose without or with increasing concentrations of insulin. After incubation, lipids were extracted and radioactive glucose incorporation into total lipid was used as an index of lipogenesis. The ability of insulin to stimulate lipogenesis at maximum effective concentration was expressed as nmol glucose/107 fat cells*2hrs.
Adiponectin secretion from intact adipose tissue was determined in conditioned media (100mg tissue/1ml of medium) as described before68 except that an ELISA assay from Mercodia (Uppsala, Sweden) was used for the analyses69. Results were expressed as ng/107 fat cells*2 hrs.
Values are presented as means±S.E.M. Statistical evaluation was performed with the Mann–Whitney test using the GraphPad Prism software, version 5.0 (GraphPad Software, San Diego, CA, USA) or multiple regression using JMP (12.1 SAS Institute Inc., Cary, NC, USA). Non-normally distributed parameters were log10 transformed as indicated. Asterisks or pound signs denote statistically significant differences between the indicated groups of data (n.s. no significant; * or #P<0.05; ** or ##P<0.01; *** or ###P<0.001).
How to cite this article: Peng, X. et al. Thioredoxin reductase 1 suppresses adipocyte differentiation and insulin responsiveness. Sci. Rep. 6, 28080; doi: 10.1038/srep28080 (2016).
This study was supported by funding to ESJA from Karolinska Institutet, The Swedish Research Council, The Swedish Cancer Society, to MR from the Strategic Research Program in Diabetes and to ACG from Diabetesfonden and a “Ramón y Cajal” fellowship (RYC-2014-15792) from Spanish Ministerio de Economía y Competitividad. Assistance in glucose measurements by Dr. Julie Massart, Molecular Medicine and Surgery, Karolinska Institutet, and constructive comments on the manuscript by Prof. Ed Schmidt, Montana State University, are thankfully acknowledged.
Author Contributions X.P. and E.S.J.A. conceived the study and wrote the paper. X.P. performed the major parts of the experiments. M.C. contributed the MEFs and helped designing experiments. A.G.-C. designed and performed experiments for the results in Figure 1. P.P. and M.R. planned, conducted and analyzed the experiments in human adipocytes. M.R. analyzed microarray data from human adipose tissue. All authors contributed to interpretations of results and to writing of the final manuscript.