Thioredoxin-interacting Protein Mediates High Glucose-induced Reactive Oxygen Species Generation by Mitochondria and the NADPH Oxidase, Nox4, in Mesangial Cells*

Background: Thioredoxin-interacting protein (TxNIP) is up-regulated by high glucose (HG), inhibits the antioxidant, thioredoxin, and thereby is implicated in oxidative stress. Results: TxNIP deficiency protects mesangial cells from HG-induced oxidative stress and increased collagen by blocking mitochondrial glucose metabolism, NADPH oxidase, and Nox4. Conclusion: TxNIP controls ROS generation by regulating the TCA cycle versus glycolytic glucose flux. Significance: Inhibition of TxNIP is a promising approach to treat glucose toxicity. Thioredoxin-interacting protein (TxNIP) is up-regulated by high glucose and is associated with oxidative stress. It has been implicated in hyperglycemia-induced β-cell dysfunction and apoptosis. As high glucose and oxidative stress mediate diabetic nephropathy (DN), the contribution of TxNIP was investigated in renal mesangial cell reactive oxygen species (ROS) generation and collagen synthesis. To determine the role of TxNIP, mouse mesangial cells (MC) cultured from wild-type C3H and TxNIP-deficient Hcb-19 mice were incubated in HG. Confocal microscopy was used to measure total and mitochondrial ROS production (DCF and MitoSOX) and collagen IV. Trx and NADPH oxidase activities were assayed and NADPH oxidase isoforms, Nox2 and Nox4, and antioxidant enzymes were determined by immunoblotting. C3H MC exposed to HG elicited a significant increase in cellular and mitochondrial ROS as well as Nox4 protein expression and NADPH oxidase activation, whereas Hcb-19 MC showed no response. Trx activity was attenuated by HG only in C3H MC. These defects in Hcb-19 MC were not due to increased antioxidant enzymes or scavenging of ROS, but associated with decreased ROS generation. Adenovirus-mediated overexpression of TxNIP in Hcb-19 MC and TxNIP knockdown with siRNA in C3H confirmed the specific role of TxNIP. Collagen IV accumulation in HG was markedly reduced in Hcb-19 cells. TxNIP is a critical component of the HG-ROS signaling pathway, required for the induction of mitochondrial and total cell ROS and the NADPH oxidase isoform, Nox4. TxNIP is a potential target to prevent DN.


Thioredoxin-interacting protein (TxNIP) is up-regulated by high glucose and is associated with oxidative stress. It has been implicated in hyperglycemia-induced ␤-cell dysfunction and apoptosis. As high glucose and oxidative stress mediate diabetic nephropathy (DN), the contribution of TxNIP was investigated in renal mesangial cell reactive oxygen species (ROS) generation and collagen synthesis.
To determine the role of TxNIP, mouse mesangial cells (MC) cultured from wild-type C3H and TxNIPdeficient Hcb-19 mice were incubated in HG. Confocal microscopy was used to measure total and mitochondrial ROS production (DCF and MitoSOX) and collagen IV. Trx and NADPH oxidase activities were assayed and NADPH oxidase isoforms, Nox2 and Nox4, and antioxidant enzymes were determined by immunoblotting. C3H MC exposed to HG elicited a significant increase in cellular and mitochondrial ROS as well as Nox4 protein expression and NADPH oxidase activation, whereas Hcb-19 MC showed no response. Trx activity was attenuated by HG only in C3H MC. These defects in Hcb-19 MC were not due to increased antioxidant enzymes or scavenging of ROS, but associated with decreased ROS generation. Adenovirus-mediated overexpression of TxNIP in Hcb-19 MC and TxNIP knockdown with siRNA in C3H confirmed the specific role of TxNIP.

Collagen IV accumulation in HG was markedly reduced in Hcb-19 cells. TxNIP is a critical component of the HG-ROS signaling pathway, required for the induction of mitochondrial and total cell ROS and the NADPH oxidase isoform, Nox4. TxNIP is a potential target to prevent DN.
Diabetic nephropathy (DN), 4 a microvascular complication of diabetes mellitus, is a major cause of morbidity and mortality resulting in end stage renal disease requiring dialysis and/or transplantation (1)(2)(3). Although chronic exposure to elevated levels of glucose is the central cause, the molecular pathogenesis remains to be completely defined (4). The pathological hallmarks of DN include increased mesangial matrix expansion associated with increased production and decreased degradation of extracellular matrix (ECM) proteins such as fibronectin, laminin, and collagen (2,(5)(6)(7)(8).
A major link between high glucose and cellular dysfunction is oxidative stress (9,10). Thus, it has been proposed that increased metabolic flux of glucose via mitochondrial glucose oxidation leads to the increased production of reactive oxygen species (ROS), such as O 2 . (superoxide), a byproduct of electron transport (9). ROS generation in excess of endogenous antioxidant neutralizing capacity leads to oxidative stress. This in turn leads to an inhibition of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) activity, promoting increased flux through upstream glycolytic branch pathways, including the aldose reductase/polyol pathway, formation of reactive sugars and advanced glycation end products, de novo synthesis of diacylglycerol and chronic activation of PKCs, and increased intracellular O-glycosylation via the hexosamine biosynthetic pathway * This work was supported in part by Canadian Institutes of Health Research (9,10). At the same time, ROS have been shown to be increased in the presence of high glucose by activation of NADPH oxidases (11)(12)(13)(14)(15)(16). In vivo studies of diabetic rodents demonstrate protection against complications by antioxidants, by inhibition of NADPH oxidase, and by genetic overexpression of antioxidant enzymes, such as Cu,Zn-superoxide dismutase (SOD) (17)(18)(19). More recently, in vivo and in vitro evidence for up-regulation by high glucose of NADPH oxidase subunits, e.g. p22 phox and p47 phox , as well as the predominant renal isoform, Nox4, strongly support this concept (13)(14)(15)(16)20).
The importance of oxidative stress as a mediator of high glucose-induced pathology has led to studies of its regulation. One protein found to be markedly up-regulated by high glucose in a number of cells, including mesangial cells and pancreatic ␤-cells is thioredoxin-interacting protein (TxNIP), also known as vitamin D up-regulated protein 1 (VDUP1) (7,(21)(22)(23)(24). TxNIP has been implicated in promoting oxidative stress by binding and inhibiting thioredoxins (Trx)-1 and -2, ubiquitous antioxidant oxidase-reductase enzymes localized to the cytosol and mitochondria, respectively (21,(25)(26)(27). Thus, by impairing ROS scavenging in high glucose TxNIP may contribute to "glucose toxicity." This has been observed in pancreatic ␤-cells as we and others have reported that TxNIP-deficient mice are relatively protected from the development of ␤-cell failure and diabetes induced by streptozotocin (28 -30) and in the presence of insulin resistance (31). A similar role in promoting oxidative stress would implicate TxNIP in the pathogenesis of DN. To investigate this possibility and explore TxNIP action directly in the kidney, primary mesangial cells were cultured from wildtype control (C3H) and TxNIP-deficient (Hcb-19) mice and exposed to high glucose. TxNIP deficiency protected the mesangial cells (MC) from high glucose-induced increased ROS and collagen accumulation. Surprisingly, the lack of TxNIP was associated with a marked decrease in NADPH oxidase activity, ROS generation, and Nox4 induction rather than an increase in ROS scavenging. Furthermore, mitochondrial O 2 . was not increased by high glucose in Hcb-19 cells, whereas lactate production was augmented. These data place TxNIP upstream of NADPH oxidase activation by high glucose and suggest a potential Trx-independent action of TxNIP to promote mitochondrial metabolic flux, oxidative stress, and DN.

MATERIALS AND METHODS
Cell Culture-Mouse MC (passages 5-12) from C3H (wildtype TxNIP) and Hcb-19 (TxNIP-deficient) mice (kindly provided by R. Davis, University of California) were isolated and characterized as described (32,33). They were propagated in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS and 1% penicillin/streptomycin. At 70 -80% confluence, cells were growth arrested with 0.5% FBS for 48 h and incubated with either 5.6 mM normal glucose (NG) or 25 mM high glucose (HG) for the times indicated. Cell lysates were obtained by homogenizing MCs in RIPA buffer containing Roche Complete protease inhibitors and passing them through a 26-gauge needle 10 times to disrupt the cells. The homogenates were centrifuged at 5000 ϫ g for 10 min at 4°C and supernatants were used immediately or stored at Ϫ80°C.  (50 nM) or TxNIP siRNA (50 nM) was mixed with polyplus reagent and serum-free Opti-MEM (Invitrogen) for 20 min at room temperature. Two hundred l were added to the C3H MC containing 1.8 ml of DMEM (10% FBS) and then incubated for 24 h before growth arrest. The recombinant adenoviruses expressing green fluorescent protein (AdGFP) and TxNIP (Ad-TxNIP) were kindly provided by Dr. R. T. Lee (Harvard, Boston, MA). These viruses were amplified in 293A cells, purified, and concentrated using the Vivapure AdenoPACK100 kit (Cedarlane). Experiments were conducted using stock titer of 10 9 infectious units (ifu)/ml. Briefly, a mixture containing DMEM with 15% FBS, 2.5 mg/ml of poly-L-lysine and adenovirus was added to subconfluent Hcb-19 MC and incubated for 24 h before growth arrest. After preliminary dose-response experiments demonstrating levels of protein expression by immunoblotting (data not shown), 250 l of stock in 1.75 ml of media (25 ϫ 10 7 ifu/10 6 cells) to 1000 l (10 9 ifu/10 6 cells) were chosen for these studies.

ROS and Mitochondrial Membrane Potential Measurement-
MCs were cultured on glass coverslips and incubated in the dark with 1 M carboxymethyl-H 2 -dichlorofluorescein diacetate (CM-H 2 DCFDA) or dihydroethidium (DHE) for 30 min at 37°C. Intracellular ROS production was assessed with an Olympus FluoView 1000 Laser Scanning Confocal Microscope (excitation/emission at ϭ 488 nm/515 nm for DCF; excitation/emission at 396 nm/579 nm for superoxide, and excitation/emission at 510 nm/580 nm for general ROS detection for DHE). Mitochondrial superoxide formation was detected by incubating cells in the dark with 5 M MitoSOX Red dye (excitation/emission at ϭ 510 nm/580 nm) for 30 min. To assess mitochondrial membrane potential, the cells were preincubated with 2 l/ml of JC-1 dye (Invitrogen) for 30 min and detected at excitation/emission at ϭ 590/610 nm for JC-1 aggregates and excitation/emission at ϭ 485/535 nm for monomers. The average fluorescence intensity per cell for each experimental group of cells was calculated using NIH ImageJ analysis software. An average of 10 cells/field and at least 4 -5 random fields/condition were chosen for each experiment and analyzed in a blinded manner. Hydrogen peroxide released by the cells was assessed by incubating the growth media (phenol red-free) from the cells in a reaction mixture containing Amplex Red (Invitrogen) and then determining the absorbance at 560 nm in a microplate reader, as per the manufacturer's instructions. The readings were then normalized by the number of cells for each condition.
The rate of hydrogen peroxide disappearance in MCs was assessed according to Chen et al. (40) with a slight modification. Briefly, growth-arrested cells were pretreated with 10 M diphenyleneiodonium chloride (Sigma) for 1 h to inhibit endogenous NADPH oxidase and mitochondria-derived ROS (37). Then different concentrations of H 2 O 2 (hydrogen peroxide) (1 M-100 M) were added to the cells and the changes in the levels of ROS were measured at various intervals using CM-H 2 DCFDA dye and a Molecular Device plate reader at excitation/emission at ϭ 485/538 nm. Based on the results from preliminary experiments, the rates of ROS decomposition were linear between 120 and 180 min post-H 2 O 2 treatment; thus, the rates of ROS degradation under the various conditions were determined during this 1 h.
Mitochondria Isolation-MCs were harvested from the cultures and mitochondria were isolated using a ThermoFisher mitochondria isolation kit according to the manufacturer's instructions.
Confocal Imaging of Collagen ␣ (Type IV)-MCs were cultured on glass coverslips and treated with NG or HG for 24 h after growth arrest. Cells were fixed in 3.7% formaldehyde, permeabilized in methanol, and blocked with 1% goat serum plus 0.1% BSA. Immunofluorescence staining was performed by incubating the fixed cells with collagen ␣ (IV) primary antibody (1:250, Rockland) and FITC-conjugated secondary antibody (1:100, BD Transduction Labs) and viewed with an Olympus FluoView Confocal Microscope.
NADPH Oxidase Activity Assay-NADPH oxidase activity was measured using the lucigenin-enhanced chemiluminescence method as previously described (11). In brief, MCs were washed and homogenized in lysis buffer containing 20 mM KH 2 PO 4 (pH 7.0), 1 mM EGTA and Complete protease inhibitors (Roche Applied Science). A total of 100 l of homogenates were added to 900 l of assay buffer comprising 50 mM KH 2 PO 4 (pH 7.0), 1 mM EGTA, 150 mM sucrose, 5 M lucigenin (electron acceptor), and 100 M NADPH (electron donor). Photoemission was measured every 30 s for 5 min with a GloMax luminometer. A buffer without NADPH was used as blank. Rate of superoxide generation was calculated by subtracting the readings from the blank and expressed as relative light units per min per mg of protein.
Trx Activity Assay-After cell lysis, thioredoxin activity assay was performed according to Holmgren and co-workers (38), with modifications. Briefly, duplicate samples of 50 g of protein (from cell lysates) were added to 40 l of mixture 1 (400 l of 1 M HEPES (pH 7.45), 160 l of 0.2 M EDTA, 120 l of NADPH (40 mg/ml), and 1000 l of insulin (10 mg/ml)), bringing the final volume to 110 l with water. To the first sample, 10 l of Trx reductase was added, and to the second sample, 10 l of water. The mixtures were incubated at 37°C for 20 min and the enzymatic activity was stopped by adding 250 l of stop buffer (6 M guanidine HCl and 0.4 mg/ml of 5,5Ј-dithiobis(nitrobenzoic acid)) to each tube. Two hundred l from each tube was added to 96-well plates and readings were taken at 410 nm. The differences in the absorbance between Trx reductase and water-treated samples represent the thioredoxin activity in the cell lysates.
Lactate Concentrations-Lactate concentrations in the cell culture media were assessed using the Abcam Lactate Colorimetric Assay kit according to the manufacturer's instruction.
Statistical Analysis-Results are expressed as mean Ϯ S.E. Statistical analysis was performed by analysis of variance followed by the Newman-Keuls method for multiple comparisons using GraphPad Prism 4. Comparison between two sets of samples was analyzed by Student's t test. p Ͻ 0.05 was considered to be statistically significant.

High Glucose Induces Expression of Collagen IV in Wild-type C3H but Not in TxNIP-deficient Hcb-19 Mesangial Cells-
There is evidence in the literature demonstrating the effect of high glucose on the expression of TxNIP (39 -41). In this study we verified that the primary mouse MCs harvested from C3H control mice exhibited a similar response. However, in Hcb-19 MC, which lack TxNIP protein due to a spontaneous nonsense mutation in the gene encoding TxNIP (42), showed no expression by immunoblotting (Fig. 1A). Thus, treatment with high glucose resulted in a time-dependent increase in TxNIP, which was significantly increased in wild-type C3H MC by 1 h of HG treatment, rising to ϳ15-fold by 24 h (Fig. 1A).
One of the hallmarks of DN is increased mesangial ECM protein deposition (2,43). Therefore, we determined whether TxNIP was required for this outcome. Collagen ␣(IV), an ECM protein known to be associated with glomerulosclerosis was increased 1.7 Ϯ 0.43-fold after a 24-h HG treatment of C3H MC, but no change was observed in Hcb-19 MC (Fig. 1B). These data, consistent with Kobayashi et al. (7), indicate that TxNIP is required for the induction of collagen IV protein by high glucose in MC.
TxNIP Is Required for the High Glucose-induced Increase in MC ROS and Decrease in Trx Activity-Intracellular ROS production was assessed by preincubating mesangial cells with a cell-permeable fluorescent probe, CM-H 2 DCFDA, which is oxidized to 2Ј7Ј-dichlorofluorescein by H 2 O 2. As illustrated in Fig. 2A, HG-induced ROS production was augmented in C3H cells, peaking at 3 h of treatment. However, HG exposure did not increase ROS in the Hcb-19 cells. Similar experiments were performed using DHE, another biomarker for cellular ROS. Fluorescent detection of superoxide production by the MC (Fig. 2B) was assessed at excitation/emission at 396/579 nm, and total ROS (Fig. 2C) that forms ethidium at 510/580 nm, as described (44). The pattern of superoxide and total ROS generation were similar to the results obtained from the DCF experiments (Fig. 2, B and C). To confirm that ROS signaling in C3H and Hcb19 cells (Fig. 2, A-C) were not due to differences in the uptake of the dye by the different cells, the amount of H 2 O 2 released was also assessed by the levels of resorufin (red fluorescent-oxidized product of Amplex Red reagent) in the cell culture media. As seen in Fig. 2D, the amount of H 2 O 2 generated by the cells was similar to the ROS detected by DCF/DHE (Fig. 2, A-C). Together, these data indicate that without TxNIP, there appears to be either an impairment in ROS generation or an increase in endogenous ROS scavenging/antioxidant capacity.
To gain insight into the extent of Trx activity available to scavenge intracellular ROS, a Trx activity assay was performed using a modified version of the protocol of Holmgren and coworkers (38). Interestingly, there was a significant reduction in the Trx activity in C3H MC by 1 h of HG treatment that was decreased by 69% (versus C3H NG) at 3 h post-treatment (Fig.  2E). This decline in Trx activity, representing oxidation/inactivation of endogenous Trx by the ROS and/or inhibition by TxNIP, could have partially accounted for the observed increase in ROS (Fig. 2, A-D). On the other hand, although there was a trend toward a decrease in Trx activity in high glucose in Hcb-19 cells, the difference was not statistically significant.
H 2 O 2 Scavenging Is Not Augmented in TxNIP-deficient Mesangial Cells-The lack of a significant decrease in endogenous Trx activity in TxNIP-deficient cells in the presence of high glucose was surprising and suggested that either there existed such an excess of antioxidant activity that a decrease was not detected, or that ROS generation was impaired. To more directly determine the ability of Hcb-19 MC to dissipate ROS, cells were exposed to 1 and 10 M H 2 O 2 and ROS disappearance was monitored over time by DCF fluorescence as described under "Materials and Methods." The rate of disappearance of ROS was lower in Hcb-19 compared with C3H MC (Fig. 2F). These results appeared paradoxical because TxNIP is known to inhibit Trx and thus antioxidant activity. However, whereas Trx1 and -2 protein levels were minimally decreased in the basal state in Hcb-19 cells, they were rapidly up-regulated (3 h) in high glucose in C3H but not altered in Hcb-19 MC (Fig. 3). Together, these data suggested a defect in ROS generation in Hcb-19 cells.
High Glucose-stimulated NADPH Oxidase Activity and Nox4 Expression Are Impaired in Hcb-19 Cells-A significant source of high glucose-stimulated ROS is NADPH oxidase. To investigate its role in the decreased ROS generation in TxNIP deficiency, NADPH oxidase activity was measured before and after various times of high glucose exposure. Superoxide generation by the NADPH oxidase system was assessed by the lucigeninenhanced chemiluminescence method in total cell lysates from MC treated with HG for 3 and 24 h. NADPH oxidase activity was decreased in the basal state in the absence of TxNIP in Hcb-19 MC (Fig. 4A). After 3 h of HG exposure there was a 2.3 Ϯ 0.16-fold increase in superoxide production in C3H MC that remained higher at 24 h. In contrast, there were no significant changes in superoxide production in Hcb-19 cells after 3 or 24 h of HG (Fig. 4A). The two major isoforms of NADPH oxidase, which are expressed in kidney, are Nox2 and Nox4 with a small amount of Nox1 (4,45). Indeed, Nox4, also known as "renox," is the most highly expressed. To investigate the contribution of Nox isoforms in the HG-induced NADPH oxidase activity, we determined their levels in the two cell types. Nox1 was only weakly detected, showed no difference between C3H and Hcb-19 MC, and was not up-regulated by HG (Fig. 5A). Nox2 was readily detected but also not different. We observed a trend toward increased Nox2 in HG in C3H MC but this did not reach statistical significance even after 24 h (Fig. 4B). In addition, Nox2 mRNA was not altered by HG in either C3H or Hcb-19 MC (Fig. 4D). In contrast, Nox4 protein was lower in TxNIP-deficient Hcb-19 MC in the basal state and not altered by HG, whereas in C3H Nox4 was rapidly induced (3 h) and remained elevated at 24 h (Fig. 4C). Similar results were obtained by real-time RT-PCR assessment of Nox4 mRNA, which was increased by HG only in C3H MC (Fig. 4E). The increase in NADPH oxidase activity stimulated by HG has also been associated with up-regulation of other Nox-associated subunits such as p47 phox , p67 phox , rac1, and p22 phox (4,13,15,46,47). Although p22 phox could contribute to Nox2 and/or Nox4, the former 3 proteins are associated with Nox2 activity (4,45,48,49). Under these conditions there was a significant but transient increase in p47 phox induced by HG in C3H, which returned to basal levels after 24 h. However, there was no change detected in Hcb-19 (Fig. 5B). Similarly, there were no changes caused by HG in p67 phox or rac1 in either cell type (Fig.  5, C and D). p22 phox was also significantly induced by HG only in C3H MC, and this increase was maintained at 24 h (Fig. 5E). Thus, whereas we cannot rule out a contribution to the HGstimulated NADPH oxidase activity of Nox2 at 3 h, these data suggest that Nox4 is the major isoform involved, particularly at 24 h. In addition, TxNIP is necessary for up-regulation of Nox4 as well as the subunits, p47 phox and p22 phox in HG.
High Glucose Promotes ROS Production and Nox4 Protein Expression in the Mitochondria of C3H MC-There are studies that have demonstrated that Nox4 siRNA or antisense oligonucleotides block high glucose-induced increases in mesangial cell ROS, i.e. cytosolic NADPH oxidase activity as well as mitochondrial superoxide formation, indicating that Nox4 is essential for HG-induced MC ROS generation (14,16). These findings are also consistent with the observation that Nox4 is an important mediator of mesangial matrix accumulation in diabetes (14). To examine mitochondria-specific ROS generation and Nox4 expression, MC were treated with HG for varying times and superoxide generation in live cells was measured with MitoSOX Red, a mitochondria-specific fluorogenic dye that produces red fluorescence upon being oxidized by superoxide. By 1 h of HG treatment, there was a significant rise in mitochondrial superoxide production in C3H MC, which remained elevated up to 24 h of HG treatment (Fig. 6A). However, mitochondrial ROS production by Hcb-19 MC was not affected by HG (Fig. 6A), indicating mitochondrial dysfunction, at least in this respect, in Hcb-19 cells lacking TxNIP.
In MC, Nox4 has been found to primarily localize to mitochondria and to a lesser extent to plasma membranes (16).
Thus, the expression of Nox4 in mitochondrial extracts from HG-treated MCs was examined. The mitochondrial fractions were prepared as described under "Materials and Methods" and purity was verified by immunoblotting of the mitochondrial markers prohibitin, COX IV and VDAC/porin (Fig. 6B). As shown in Fig. 6C, there was a dramatic increase in Nox4 levels in HG-treated C3H cells, but in Hcb-19, mitochondrial Nox4 expression remained unaffected. Therefore, it appears that TxNIP plays an important role in inducing mitochondrial Nox4, which in turn contributes to the ROS production.
Knockdown of TxNIP in Wild-type C3H MC Mimics the Features of Hcb-19 Cells-To verify that the defects in ROS generation, NADPH oxidase activation and Nox4 expression upon HG exposure in TxNIP-deficient Hcb-19 MC were caused by the lack of TxNIP, the wild-type C3H cells were treated with TxNIP-specific or scrambled siRNA. TxNIP protein was effectively decreased upon knockdown with siRNA in HG to levels even below that of scrambled siRNA-transfected cells and completely blocked the up-regulation of TxNIP by HG (Fig. 7D). This decrease in TxNIP markedly inhibited HG-stimulated cellular ROS monitored by DCF fluorescence (Fig. 7A), mitochondria-specific ROS detected with MitoSox (Fig. 7B), and NADPH oxidase activity (Fig. 7C). In addition, TxNIP knockdown blocked Nox4 induction (Fig. 7D) in response to HG and collagen IV accumulation, an in vitro marker of the increased ECM protein in DN (Fig. 7E). These results in C3H cells with TxNIP-specific knockdown indicate the requirement of TxNIP as a mediator of HG signaling.
Overexpression of TxNIP in Hcb-19 MC Augments Mitochondrial ROS and Nox4-To confirm the functional role of TxNIP in HG signaling, the TxNIP-deficient Hcb-19 MC were transduced with TxNIP expressing adenovirus (AdTxNIP) or control GFP expressing adenovirus (AdGFP). Dose-response experiments revealed that 25 ϫ 10 7 ifu/ml/well (10 6 cells) for 24 h resulted in TxNIP protein levels that were at least as high as  (Fig. 8A). However, AdTx-NIP-transfected Hcb-19 cells showed increased MitoSox fluorescence in normal glucose (ϳ2-fold of AdGFP Hcb-19), which did not increase further with HG treatment (Fig. 8A). Measurement of Nox4 protein showed a TxNIP dose-dependent increase in Hcb-19 cells in normal glucose, which also did not increase further in HG (Fig. 8B). These data indicate that TxNIP, when expressed at levels similar to that observed with HG in MC, induces mitochondrial ROS and Nox4 expression and support the notion that it plays a key role in diabetic nephropathy.
TxNIP Is Required for the High Glucose-associated Increase in Mitochondrial/TCA cycle Glucose Metabolism-TxNIP has been recently implicated as a "gatekeeper" of pyruvate conversion to acetyl-CoA for mitochondrial glucose oxidation via the TCA cycle and electron transport (50). Cardiomyocytes from TxNIP Ϫ/Ϫ mice display increased glycolysis and lactate production, a Warburg-like effect (50). Evidence for this metabolic pattern was observed in Hcb-19 MC in which exposure to HG resulted in a significant elevation of lactate concentrations in the medium that was not observed in C3H (Fig. 9A).
Glucose metabolite flux and oxidation in mitochondria were assessed by measuring the mitochondrial membrane potential. Mitochondrial membrane potential is increased by flux through the electron transport chain and generation of a proton gradient across the membrane (51). JC-1 is a cationic dye that selectively enters and accumulates in mitochondrial matrix, where it forms J-aggregates with intense red fluorescence upon exceeding critical concentrations, thereby changing its color from green to red as the membrane potential is increased (52). In control C3H MC, mitochondrial membrane potential increased when HG was added forming red JC-1 aggregates (Fig. 9B, upper panel). In contrast, in Hcb-19 cells, JC-1 remained in its monomeric form as shown by green fluorescence (Fig. 9B, lower panel), indicating that mitochondrial glucose metabolism was not augmented in these cells. These results support the concept that TxNIP plays a key role augmenting glucose metabolic flux via the TCA cycle and that the defects in HG signaling may be mediated, at least in part, by this deficiency in mitochondrial function.

DISCUSSION
The importance of elevated levels of ROS in the pathogenesis of the microvascular complications of diabetes has been well documented (11,12,53). Although increased mitochondrial glucose oxidation and flux through the electron transport chain have been suggested to be the primary source, elevated NADPH oxidase activity has also been demonstrated to contribute to increased ROS and ultimately, to profibrotic signaling in the kidney stimulated by hyperglycemia (11)(12)(13)(14)(15). The cellular levels of ROS and the presence of oxidative stress are determined not only by rates of ROS generation, but also their neutralization and degradation by endogenous antioxidants. It has been pro-posed that by binding to and inhibiting the antioxidant Trxs, TxNIP up-regulation by HG promotes oxidative stress (27,54). A specific role in the kidney was suggested by Kobayashi et al. (7) who showed that overexpression of TxNIP in cultured MC led to increased collagen IV mRNA and protein. A truncated TxNIP mutant that failed to bind Trx lacked this effect (7). In this study, we found a rapid, marked and sustained up-regulation of TxNIP in primary cultures of wild-type C3H MC exposed to HG. This was associated with increased production of ROS and collagen accumulation. In contrast, in Hcb-19 TxNIP-deficient MC, exposure to HG did not stimulate ROS or increase collagen. We reasoned that the inhibition of Trx in HG by augmented TxNIP would limit the dissipation of ROS and promote oxidative stress and that an absence of TxNIP would or TxNIP (AdTxNIP) adenoviruses and C3H cells with AdGFP alone to serve as a control. After 24 h, cells were growth arrested for 48 h and then exposed to NG (5.6 mM) or HG (25 mM) for 3 h. A, mitochondrial superoxide production in the MC was determined with MitoSOX dye by confocal microscopy. Quantification is illustrated in the graph. Results are mean Ϯ S.E. *, p Ͻ 0.05; and **, p Ͻ 0.001 versus AdGFP C3H NG. B, immunoblotting for Nox4 and TxNIP was performed using lysates from Hcb19 MC transduced with varying concentrations of AdTXNIP (25 ϫ 10 7 , 5 ϫ 10 8 , and 1 ϫ 10 9 ifu/ml) in NG or HG. Results are mean Ϯ S.E. (n ϭ 4). *, p Ͻ 0.05 versus AdGFP Hcb19 NG. permit a more effective antioxidant neutralizing capacity. However, in the Hcb-19 MC the TxNIP deficiency was associated with a lack of response to HG of NADPH oxidase activity and up-regulation of the major renal NADPH oxidase isoform, Nox4. These data indicated a defect in ROS generation. Indeed, direct assessment of H 2 O 2 dissipation in the two MC lines revealed that in contrast to what was expected, ROS degradation was decreased in the absence of TxNIP. In the Hcb-19 MC, the protein content of several endogenous antioxidant enzymes including Trx1, Trx2, catalase, and HO-1 (hemoxygenase-1) tended to be lower in the basal state and showed no increase in HG in contrast to C3H MC (Fig. 3). Although we cannot rule out a contribution of Nox2 to the HG-induced NADPH oxidase activity in C3H MC, the lack of up-regulation of p47 phox and p22 phox in Hcb-19 MC (Fig. 5) would also be consistent with an absent NADPH oxidase response to HG without TxNIP.
The mechanism of up-regulation of Nox4 by HG is not clear. However, there are multiple stimulatory transcriptional regulators including NFB, HIF1-␣, AP-1, and Nrf-2 (55)(56)(57)(58)(59), which may all be activated by HG (10,11,60). Under basal conditions, Trx in mitochondria binds and inhibits ASK (apoptosis signalregulating kinase)-1, an upstream MKK3/6 and p38 MAPK activator (61)(62)(63). TxNIP up-regulation has been proposed to bind Trx and release ASK-1 to activate p38 (64 -66), which in turn would stimulate AP-1 (67,68). In support of this possibility, TxNIP has recently been reported to translocate to the mitochondria in the presence of ROS in pancreatic ␤-cells (65). It has also been found that NFB transcriptional activation requires ROS (69) and NFB is activated in response to HG in endothelial cells (70). Recently, the activation of Nrf-2 in response to the increased ROS induced by HG has been documented and appears to be a negative feedback, protective pathway by inducing an antioxidant response (71). The lack of upregulation of enzymes such as HO-1, a Nrf-2 responsive gene, in Hcb-19 MC in HG is consistent with the absence of ROS generation.
Because ROS are well documented cell signaling molecules, it is not clear whether the defect(s) observed in NADPH oxidase activation is primary or secondary. In this context, we examined mitochondria-specific ROS generation using MitoSox. In contrast to controls, in the absence of TxNIP, there was no increase in mitochondrial ROS (O 2 . ) in response to HG in Hcb-19 MC. The key role of TxNIP was demonstrated by reproducing these defects in wild-type C3H MC using TxNIPspecific siRNA-mediated knockdown (Fig. 7). Thus, decreased total cellular ROS and mitochondrial O 2 . was associated with decreased NADPH oxidase activity, Nox4, and collagen expression, were documented in response to HG in C3H MC with TxNIP knockdown compared with C3H MC transfected with scrambled siRNA. Furthermore, restoration of TxNIP by adenoviral expression induced mitochondrial ROS and Nox4 expression in Hcb-19 cells even in the absence of HG. Increased mitochondria-derived ROS in response to HG has been proposed to act as an initiator of the alterations in glucose metabolic pathways that contribute to the complications of diabetes (9 -11). Thus, a relevant action of TxNIP could be its recently appreciated function to regulate glucose metabolism, specifically to increase mitochondrial glucose oxidation (50). In cardiomyocytes, protection from ischemia-reperfusion injury in intact hearts from TxNIP Ϫ/Ϫ mice was associated with increased glycolysis to lactate and a defect in entry of pyruvate into the Krebs cycle (50). In HG, increased mitochondrial glucose oxidation is observed and we found a marked increase in the mitochondrial membrane potential in C3H MC exposed to HG. However, this increase was completely absent in TxNIPdeficient Hcb-19 MC (Fig. 9) and was accompanied by an increase in lactate. These data in MC are similar to the glucose metabolic phenotype reported in cardiomyocytes and are consistent with a primary mitochondrial defect in ROS generation. It is not clear whether this metabolic effect is mediated by the Trx binding site of TxNIP or possibly, by its ␣-arrestin domain (72).
In summary, this study demonstrates widespread defects in ROS generation in response to HG in MC, which lack TxNIP, from both mitochondria and NADPH oxidase(s). Although TxNIP inhibits the endogenous antioxidant Trx, the decrease in ROS is not due to more rapid degradation of these reactive molecules, but appears to be mediated, at least in part, by a defect in regulating glucose metabolic flux to the TCA cycle. Although we cannot exclude other direct actions of TxNIP, which binds various proteins (73)(74)(75), this defect in mitochondrial and Nox4-mediated ROS generation would inhibit multiple signaling pathways such as PKC, advanced glycation end product formation, O-glycosylation, and TGF-␤1, activated by high glucose. Thus, the lack of TxNIP was associated with protection from the pro-fibrotic signaling to collagen expression by HG. In view of the marked up-regulation of TxNIP and the key signaling role of ROS in HG-induced microvascular complications, TxNIP may be a relevant and powerful novel therapeutic target in their treatment and prevention.