Subcellular localization of Nox4 and regulation in diabetes

Several approaches were used to confirm Nox4 localization. First, crude subcellular fractions were prepared from rat kidney cortex homogenates (12). Immunoblot analysis using the rabbit polyclonal anti-Nox4 antibodies shows that a 70- to 75-kDa band corresponding to Nox4 localizes in both crude membrane (Mb) and mitochondrial fractions (Mit) (Fig. 1A Upper). Note that in some mitochondrial preparations, we observed multiple Nox4-immunoreactive bands (Fig. 1A Lower). Similar results were obtained by using a commercial Nox4 antibody raised to an epitope of the amino terminus of Nox4 (Fig. S2A). The two Nox4 antibodies were generated against different epitopes. Importantly, in the Mit fraction, both antibodies recognize predominant bands at 70–75 kDa. Detection of multiple bands in the Mit fraction is likely due to processing of proteins imported to the mitochondria such as clipping, oxidation or sumoylation that can affect their electrophoretic mobility (13–16). The detection of multiple bands by Nox4 antibody in the total fractions is consistent with the existence of isoforms of Nox4 previously reported by Goyal et al. (17). However, further characterization of these variants is needed. The rabbit antibody generated in the laboratory predominantly recognizes the 70- to 75-kDa band that is also detected by the commercial antibody. In Fig. 1B, the Nox4 antibody was used to analyze Nox4 protein expression in fractions of rat renal cortex membranes prepared at different gravitational forces as reported (18). Nox4 is predominantly present in the 1,000, 3,000, and 7,000 × g, heavy Mit fractions. Immunoblotting using prohibitin antibodies, a mitochondrial marker, confirmed that these fractions are enriched in mitochondria (Fig. 1B). Nox4, as expected, was also detected to a lesser extent in the 17,000 × g pellet, a fraction enriched in plasma membranes (Fig. 1B). Prohibitin was not detected in the pellet, excluding mitochondrial contamination of this fraction. Pure mitochondria were also prepared from rat kidney cortex by using a combination of differential and Percoll gradient centrifugation (19). The brown mitochondrial band was directly collected, and the purity of the preparation was confirmed by electron microscopy, where mostly intact mitochondria are seen (Fig. 1C). To further demonstrate that Nox4 is localized within the mitochondria, Percoll gradient-purified mitochondria were resolved by SDS/PAGE and probed for Nox4. Western blot analyses showed that mitochondrial Nox4 expression is robust (Fig. 1D). The purity of mitochondria from kidney cortex was determined by the expression of prohibitin (mitochondria) and the absence of Na⁺/K⁺-ATPase (plasma membrane), calnexin (endoplasmic reticulum), ADP ribosylation factor (ARF; Golgi), or LAMP-1 (lysosome) (Fig. 1D). Titration of the pure Mit fraction confirms that our Nox4 antibody predominantly detects the 70- to 75-kDa band (Fig. 1E). Mitochondrial Nox4 expression was verified in the total and Percoll gradient-purified mitochondria fractions by using a commercial antibody (Fig. S2B).

We also investigated the mitochondrial localization of Nox4 in rat kidney glomerular mesangial cells (MCs), vascular pericytes known to express an active Nox4-based superoxide generating NAD(P)H oxidase (11). MCs constitute a primary target for glucose-mediated oxidative injury (11). Differential centrifugation of MC extracts yielded several subcellular fractions that were probed for Nox4. Similar to kidney cortex, Nox4 was detected as a 70- to 75-kDa protein in crude Mit and Mb fractions (Fig. 2A). The mitochondrial extract was positive for the mitochondrial-specific protein cytochrome (Cyt) c, by immunoblotting (Fig. 2A). Mitochondria were also isolated by using a Mit fractionation kit that allowed the fractionation of MCs into cytosol (C) and mitochondria (Mit) (Fig. 2B). Western blot analysis showed that in MCs, Nox4 localizes to mitochondria, but not to C (Fig. 2B), similar to the above findings in kidney cortex. As described in the studies with kidney cortex, purified mitochondria were also obtained by Percoll gradient centrifugation of the crude mitochondria fraction. In this set of experiments, the mitochondrial band was collected by gradient fractionation. Nox4 was exclusively detected in fraction 5, which also expresses the mitochondrial marker prohibitin (Fig. 2C; Fig. S3). Immunoblot analysis with increasing amounts of pure mitochondria isolated from MCs also yielded a predominant 70- to 75-kDa band (Fig. 2D). These results demonstrate that Nox4 cofractionates with purified mitochondria in MCs.

Immunofluorescence confocal microscopy was also used to localize Nox4 in MCs by using Cyanin-3-labeled or FITC-conjugated anti-rabbit IgG. Nox4 colocalizes with the mitochondrial markers Mitotracker Green FM (MTG) or Mitotracker Deep Red 633 (MTR) (Fig. 2 E and F, respectively). Some Nox4 also localizes to the plasma cell membrane. Nox4 localization to the mitochondria was independently observed by using commercially available Nox4 antibodies (Fig. 2G). In contrast, Nox4 staining had little or no overlap with a marker of endoplasmic reticulum, the protein disulfide isomerase (PDI) (Fig. 2H). The comparison of Nox4 staining (Fig. 2 E-G) with the staining of the Golgi complex marker Golgin-97 (Fig. 2I) indicates that Nox4 was not present in this organelle. Immunofluorescence analysis also showed that Nox4 staining overlaps with MTR in other renal cells, including glomerular endothelial (Fig. S4A) and epithelial cells (Fig. S4B), as well as in nonrenal cells such as aortic endothelial cells (Fig. S4C) and vascular smooth muscle cells (Fig. S4D), indicating that Nox4 mitochondrial localization was not specific to kidney cells. In summary, confocal immunofluorescence microscopy confirmed colocalization of Nox4 to mitochondria in cultured cells. It should be mentioned that Nox4 was found to localize to various subcellular compartments such as endoplasmic reticulum, focal adhesions, nucleus, and the plasma membrane (20–24). These observations may be due to differences in cell types expressing different Nox4 isoforms and/or different cellular functions that requires the association of Nox4 isoforms with signaling molecules in specific cellular domains (20, 21). However, in the present study, we consistently find that Nox4 localizes to the mitochondria in various cell types. The mitochondrial localization prediction program, MitoProt, was used to evaluate the probabilities of Nox4 mitochondrial localization. The amino acid sequences for the Nox catalytic subunits, Nox4, gp91^phox/Nox2, Nox1, and Nox3 were retrieved from the Swiss Prot database and analyzed in the web-based service MitoProt II 1.0a4. There is a strong prediction for mitochondrial localization of Nox4, but not the other Nox isoforms (Table S1). Indeed, the mitochondrial localization probability for Nox4 (97%) was identical to that of the mitochondrial protein human Cyt c oxidase subunit IV (COX IV). The cytosolic protein human tuberin was used as a negative control. We next determined whether the mitochondrial Nox4 was functional and active. NADPH-dependent superoxide generation was detected in the Percoll gradient-purified mitochondria from MCs by using lucigenin-enhanced chemiluminescence (Fig. 3). Importantly, NADPH oxidase activity was detected exclusively in fraction 5 of the Percoll gradient (Fig. S3C). The activity was inhibited by diphenyleneiodonium (DPI), an inhibitor of Nox oxidases (Fig. 3A). It has been recently proposed that Nox4 produces mostly hydrogen peroxide (23, 25, 26). Therefore, we also measured the NADPH-dependent hydrogen peroxide production in Percoll-purified mitochondria by using Amplex red. NADPH-dependent generation of hydrogen peroxide is detected, and is markedly reduced in pure mitochondria prepared from cells pretreated with DPI (Fig. 3B).

Transfection of MCs with small interference RNA against Nox4 (siNox4) resulted in down-regulation of Nox4 mRNA as examined by quantitative real-time PCR (Fig. 4A), as well as Nox4 protein expression as assessed by immunoblot analysis (Fig. 4B). Similarly, the mitochondrial staining pattern of Nox4 is strongly diminished using both our Nox4 antibody and the commercial antibody (Fig. 4C and Fig. S5, respectively). NADPH-dependent superoxide generation was significantly reduced in purified mitochondria isolated from siNox4-transfected MCs compared with scrambled siRNA (scr)-transfected cells (Fig. 4D). We also show that both independent Nox4 antibodies immunoprecipitate NADPH oxidase activity from digitonin-permeabilized mitochondria (Fig. 4E). Together, these data provide evidence that functional Nox4 oxidase is present in mitochondria. Importantly, the observation that NADPH-dependent superoxide generation can be detected in intact mitochondria supports the contention that the carboxyl-terminus of Nox4 NADPH-binding site is most likely positioned in contact with the C or the inner mitochondrial space accessible to the NADPH cofactor. The pores of the outer mitochondria membrane are large enough to allow NADPH to flow within the inner membrane space. However, the inner mitochondrial membrane is impermeable to the flow of NADPH to the inner mitochondrial matrix space. We would like to emphasize that the Percoll-purified mitochondria were intact, supporting these contentions. Intactness of the Mit fraction was demonstrated by showing that NADH oxidase activity was not altered after treatment with rotenone and morphological examination by electron microscopy (Fig. 1C; Fig. S6). Interestingly, this concept that Nox4 exists as an active enzyme is in agreement with reports indicating that the Nox4 catalytic subunit generates a significant amount of superoxide in a constitutive manner without the requirement of other membrane or cytosolic subunits essential for the activity of other Nox isoforms such as gp91^phox/Nox2 (26, 27).

Oxidative stress is believed to be a major pathogenic mechanism of diabetes complications. The nature of the enzymatic sources of oxidative stress in diabetes or on exposure of cells to high glucose (HG) is not precisely defined. Several studies have reported that overproduction of ROS by the mitochondrial electron-transport chain is responsible for hyperglycemia-induced oxidative stress and the pathogenesis of diabetic complications (2, 4–6). Brownlee (2) and Brownlee and coworkers (5) proposed that this process represents a unified mechanism of diabetic complications. However, recent studies indicate that phagocyte-like NAD(P)H oxidases of the Nox family are also a source of ROS in diabetes (11, 28–30). Glomerular MCs are an important target of HG-induced oxidative damage in diabetes. In MCs, HG elicits ROS generation via the mitochondrial electron-transport chain (6) and a Nox4-based NAD(P)H oxidase (11). These observations prompted us to investigate the role of mitochondrial Nox4 in these processes. We found that exposure of MCs to HG (25 mM d-glucose) up-regulates Nox4 protein expression (Fig. 5A) and increases NADPH-dependent superoxide generation in the total cell homogenate and in the purified Mit fraction (Fig. 5B). Studies using MitoSOX Red, a fluorogenic dye detecting selectively mitochondrial superoxide, showed that MitoSOX fluorescence is increased after exposure of MCs to HG (25 mM d-glucose) as compared with cells incubated with normal glucose concentration (NG, 5 mM d-glucose) (Fig. 5C). Of interest is that HG-stimulated mitochondrial superoxide generation colocalizes with Nox4. Down-regulation of Nox4 by transfection of the cells with siNox4 prevented the HG-induced increase in mitochondrial superoxide assessed by MitoSOX Red (Fig. 5C). Incubation of MCs with the osmotic control, mannitol (20 mM), had no effect on Nox4 protein expression and mitochondrial superoxide generation (Fig. S7). These findings suggest that mitochondrial Nox4 participates in ROS generation induced by HG. We previously reported that renal Nox4 is up-regulated in streptozotocin-induced diabetic rats (11). Here, we show that, in addition to total cortical homogenate, Nox4 protein expression was also increased in crude and Percoll-purified mitochondria isolated from diabetic kidney cortex compared with mitochondria isolated from nondiabetic controls (Fig. 6 A and B, respectively). The up-regulation of mitochondrial Nox4 protein expression in the diabetic animals was verified by using an independent commercial Nox4 antibody (Fig. S8). Porin was used as a mitochondrial marker. NADPH-dependent superoxide generation was examined in parallel by using the above subcellular fractions from control and diabetic rats. NADPH-dependent superoxide generation was globally increased in the total and pure Mit fractions of the diabetic animals (Fig. 6C). This increase in superoxide generation detected in the Mit fraction correlated with the increased mitochondrial Nox4 protein expression. Together, our data suggest that mitochondrial Nox4 contributes to the increase in NADPH oxidase activity in diabetes. Because Nox4 inhibition by antisense oligonucleotides reduced diabetes-induced renal hypertrophy and fibronectin expression (11), the present findings implicate mitochondrial Nox4 in oxidative stress, tissue hypertrophy, and matrix expansion in diabetes.

Our findings may offer an explanation for the evidence in the literature reporting that both mitochondria and Nox-containing oxidases are sources of ROS in pathological states such as diabetes. Localization of Nox4 to the mitochondria suggests that a short paracrine loop may exist, by which ROS production by mitochondrial Nox4 regulates or is regulated by ROS generation by the mitochondrial respiratory chain. In addition to mitochondrial Nox4, membrane Nox4 is activated by extracellular agonists that bind cell membrane receptors. These observations are consistent with the previous findings that Nox4 contributes to angiotensin II (31) and transforming growth factor-β (32) redox signaling. Our study places Nox4 as a central mediator that controls oxidative stress that may lead to mitochondrial dysfunction and cell injury in diseases such as diabetes. As a corollary to these findings, Nox4 may be considered as a primary target for the design of new therapeutic strategies to counteract oxidant-mediated deleterious effects associated with various diseases characterized by oxidative stress.

Materials, cell culture, reagents, previously published methods, RNA interference methods, and immunoprecipitation of NADPH oxidase activity with Nox4 antibodies are described in detail in SI Methods.

We thank D.-Y. Lee and P. Hoover for technical contributions, and J. Barnes for electron microscopy and rabbit immunization. This work was supported by grants from the Juvenile Diabetes Research Foundation Regular research (Y.G. and H.E.A.); a grant-in-aid from the South Central Affiliate of theAmerican Heart Association (Y.G.); the American Diabetes Association (H.E.A.); the National Institutes of Health K01 Award DK 076923 and Grant R01 CA 131272 (K.B.); National Institutes of Health Grants DK 43988 and DK 33665 (H.E.A.); the National Institute of Diabetes and Digestive and Kidney Diseases–National Institutes of Health George O'Brien Kidney Research Center (H.E.A. and Y.G.); and the Veterans Administration (H.E.A. and K.B.).