|Home | About | Journals | Submit | Contact Us | Français|
Growing evidence suggests that NADPH oxidase (Nox)-derived reactive oxygen species (ROS) play important roles in regulating cytokine signaling. We have explored how TNF-α induction of Nox-dependent ROS influences NF-κB activation. Cellular stimulation by TNF-α induced NADPH-dependent superoxide production in the endosomal compartment, and this ROS was required for IKK-mediated activation of NF-κB. Inhibiting endocytosis reduced the ability of TNF-α to induce both NADPH-dependent endosomal superoxide and NF-κB, supporting the notion that redox-dependent signaling of the receptor occurs in the endosome. Molecular analyses demonstrated that endosomal H2O2 was critical for the recruitment of TRAF2 to the TNFR1/TRADD complex after endocytosis. Studies using both Nox2 siRNA and Nox2-knockout primary fibroblasts indicated that Nox2 was critical for TNF-α–mediated induction of endosomal superoxide. Redox-active endosomes that form after TNF-α or IL-1β induction recruit several common proteins (Rac1, Nox2, p67phox, SOD1), while also retaining specificity for ligand-activated receptor effectors. Our studies suggest that TNF-α and IL-1β signaling pathways both can use Nox2 to facilitate redox activation of their respective receptors at the endosomal level by promoting the redox-dependent recruitment of TRAFs. These studies help to explain how cellular compartmentalization of redox signals can be used to direct receptor activation from the plasma membrane. Antioxid. Redox Signal. 11, 1249–1263.
Redox regulation has been shown to be critical in a number of physiologic and pathologic processes, such as cancer, diabetes, and cardiovascular diseases (21, 62). In this context, reactive oxygen species (ROS), such as O2•− and H2O2, have been shown to be induced in cells after receptor stimulation with various ligands such as tumor necrosis factor alpha (TNF-α) (12, 39), lipopolysaccharide (LPS) (24, 51), angiotensin II (18), platelet-derived growth factor (PDGF) (56), insulin (37), epidermal growth factor (EGF) (2), and interleukin-1β (IL-1β) (6, 39). Furthermore, the induction of ROS by these ligands has been suggested to be important for activating many of these signaling pathways (2, 18, 33, 56). These findings suggest that a variety of signaling pathways are mediated by ROS.
Considering the diffusibility of ROS and the high concentrations of redox-reactive biomolecules in the cytoplasm, the mechanism by which ROS facilitate specific signals has been increasingly studied during recent years. Current evidence suggests that significant control of ROS signals are imparted by spatially restricted production at intracellular sites where redox-regulated signaling occurs. For example, NADPH oxidases are recognized to control spatially the production of ROS during cell signaling (59). Seven known NADPH oxidase catalytic subunits exist, which share similar mechanisms of generating superoxide (O2•−) by transferring an electron from NADPH to molecular oxygen (Nox1, Nox2gp91phox, Nox3, Nox4, Nox5, Duox1, Duox2) (29). Among these Nox isoforms, the phagocytic Nox2 (gp91phox) is the most widely characterized NADPH oxidase and has also been found to be expressed in a variety of nonphagocytic cell types (18, 32, 47). Activation of the Nox2 complex requires association of several proteins at the membrane, including p67phox, p47phox, and p22phox; Rac1/2, a small GTPase, also plays an important role in activating O2•− production by this complex (29).
Nox proteins have been increasingly recognized as key elements of intracellular signaling. For example, Nox2 has been demonstrated to produce endosomal O2•− required for IL-1β induction of NF-κB (32). Expression of a dominant-negative Nox4 or Nox4-siRNA attenuates insulin-stimulated H2O2 production and downstream phosphatase signaling involved in adipocyte insulin-receptor activation (37). TNF-α also has been hypothesized to use Nox2-dependent production in its activation cascades (1). Supportive evidence includes the demonstration that inhibition of Rac1 or p47phox abrogates TNF-α–stimulated H2O2 production and NF-κB activation (12, 15, 31). Additionally, Nox1 has been associated with ROS producing endosomes required for IL-1β or TNF-α induction of NF-κB in vascular smooth muscle cells (41). Despite the clear association of Nox proteins with redox signaling, the redox-dependent events that control many of these activation pathways remain largely unknown.
The present study seeks to clarify redox-dependent components of TNF-α signaling that are important for NF-κB activation. NF-κB, a ubiquitous transcription factor, has been shown to mediate the expression of genes responsible for the activation of immune and inflammatory responses (17). TNF-α is one of the most broadly studied cytokines known to activate NF-κB (3). The signal from TNF-α is mediated by two cell-surface receptors, TNFR1 and TNFR2 (38, 57). The binding of TNF-α triggers receptor activation, resulting in the recruitment of a number of cytoplasmic signaling proteins to the TNFR complex. For example, TRAF2 interacts directly with TNFR2 (50), but it is recruited to TNFR1 via its interaction with TNFR1-associated death-domain protein (TRADD) (22, 23). TRAF2 is a member of the TRAF protein family that includes six members (TRAF1-6) (10, 25) that generally act as adaptor proteins. TRAF2 and RIP recruitment to TNFR1/TRADD complexes is required for the activation of the IKK complex and NF-κB (30).
In the present study, we investigated the redox-dependent events responsible for TNF-α/TNFR1–mediated NF-κB activation. The focus of these studies was to determine how TNFR1 activation of the IKK complex is coordinated by Nox-derived ROS. In contrast to current thinking (30), we observed that a significant portion of TNF-α–mediated NF-κB activation requires endocytosis of the receptor. Endocytosis was essential for the activation of Nox2-dependent endosomal ROS required for TRAF2 recruitment to the TNFR1/TRADD complex and the activation of IKK and NF-κB. In contrast, TRADD efficiently recruited to TNF/TNFR1 at the plasma membrane in the absence of endocytosis. The recruitment of TRAF2 to the TNFR1/TRADD complex could be partially reconstituted at the plasma membrane in the absence of endocytosis by the addition of exogenous H2O2, suggesting that Nox2 is a source of endosomal H2O2 required for TNFR1 activation. Clearance of ROS from the endosomal compartment also significantly reduced both IKK and NF-κB activation after TNF-α stimulation. Through this process, the generation of O2•− by endosomal Nox2, and its conversion to H2O2, facilitates the redox-dependent formation of an active TNFR1 complex required for redox-dependent NF-κB activation. This mechanism helps clarify how cells can partition Nox-derived ROS to selectively influence receptor activation from the plasma membrane.
MCF-7 cells were infected with recombinant adenoviruses (500 particles per cell) as previously described (33); cells were used for experiments at 48h after infection. The following E1-deleted recombinant adenoviral vectors were used: (a) Ad.GPx-1, which encodes glutathione peroxidase 1 and degrades cytoplasmic H2O2 (33); (b) Ad.Dyn(DN), which encodes a dominant-negative mutant (K44A) of dynamin and inhibits endocytosis (13); (c) Ad.NF-κBLuc, which encodes an NF-κB–responsive promoter driving luciferase expression and was used to assess NF-κB transcriptional activation in vivo (51); and (d) Ad.BglII, an empty vector with no insert, as a control for viral infection (33). For NF-κB transcriptional assays using infection with two recombinant adenoviruses, a slightly modified sequential infection method was used (51). In this case, cells were infected with experimental vectors [i.e., Ad.Dyn(DN) or Ad.GPx1] 24h before infection with Ad.NF-κBLuc], and cells were used for experiments at 48h after initial infection. Transduction efficiencies with recombinant adenoviruses were typically 80–90%, as assessed by Ad.CMV-GFP reporter gene expression. siRNAs against Nox2 (32) were obtained from Santa Cruz Biotechnology (Santa Cruz, CA); transfections were performed by using methods and reagents described by the manufacturer. The sequences used for the siRNAs were proprietary and were not provided by the company.
MCF-7 cells were treated with recombinant TNF-α (0.5ng/ml) or IL-1β (1ng/ml) for 20min before vesicular isolation, whereas primary mouse dermal fibroblasts were treated with 10ng/ml TNF-α. For antioxidant enzyme endosomal loading experiments, purified bovine Cu/Zn superoxide dismutase (SOD1; Oxis Research, Foster City, CA) or catalase (Sigma-Aldrich, St. Louis, MO) proteins or both were diluted from concentrated stocks in PBS into fresh medium to a final concentration of 1mg/ml and then applied to cells 10min before cytokine treatment in the continued presence of SOD or catalase or both. Cells were washed and scraped into ice-cold phosphate-buffered saline (PBS). Cell pellets were then resuspended in 0.5ml of homogenization buffer (0.25M sucrose, 10mM triethanolamine, 1mM EDTA, 1mM phenylmethylsulfonyl fluoride, and 100μg/ml aprotinin), homogenized in a Duall tissue grinder, and centrifuged at 2,000g at 4°C for 10min. The supernatant was designated the postnuclear supernatant (PNS). The PNS was subsequently combined with a 60% Iodixanol (OptiPrep Axis-Shield, Norton, MA) solution to obtain a final concentration of 32% and then bottom loaded into an SW55Ti centrifuge tube under two-step gradients of 24% and 20% Iodixanol in homogenization buffer. Samples were centrifuged at 88,195g for 2h at 4°C. Fractions were collected from the top to the bottom of the centrifuge tube at 4°C (300μl per fraction) and used immediately for NADPH oxidase activity and immunoisolation assays or were frozen for Western blot assays.
NF-κB transcriptional activity was assessed by using the previously described NF-κB–inducible luciferase reporter vector (Ad.NF-κBLuc) (51). Luciferase activity was assessed at 6h after cytokine treatment (unless otherwise specified) by using 5μg of cell lysate. NADPH oxidase activities were analyzed by measuring the rate of·O2− generation with a chemiluminescent, lucigenin-based assay (35), as previously applied to vesicles (32). In brief, 5μM lucigenin (Sigma-Aldrich) in PBS was incubated with vesicular fractions for 10min in the dark, and reactions were initiated by addition of 100μM NADPH (Sigma-Aldrich). The change in luminescence was then measured over the course of 3min (five readings per second).
Rac1-containing endosomes were isolated based on a previous method (58). Cells were infected with Ad.HA-Rac1 48h before TNF-α treatment. After Iodixanol isolation of intracellular vesicles, one-half of the combined peak vesicular fraction was used directly for biochemical analyses, and the other half was used for immunoaffinity isolation by using Dynabeads M-500 (Dynal Bioscience, Oslo, Norway) coated with the anti-HA antibody. Before use, beads were coated with antibodies as follows: the secondary antibody (anti-rat antibody) was conjugated to Dynabeads (4×108 beads/mL) in 0.1M borate buffer (pH 9.5) for 24h at 25°C with slow rocking. The beads were then placed into the magnet for 5min and washed in 0.1% (wt/vol) bovine serum albumin (BSA)-PBS for 5min at 4°C. A final wash in 0.2M Tris (pH 8.5)-BSA was performed for 24h. Finally, the beads were resuspended in BSA-PBS and conjugated to 4μg of primary anti-HA antibody per 107 beads overnight at 4°C and then washed in BSA-PBS. Vesicular fractions were mixed with 700μl of coated beads in PBS containing 2mM EDTA, 5% BSA, and protease inhibitors. The mixture was incubated for 6h at 4°C with slow rocking, followed by magnetic capture and washing in the same tube 3 times (15min each). Beads with HA-enriched endosomes were then resuspended in PBS, and wash supernatants were saved for analysis.
Western blotting was performed by using standard protocols, and protein concentrations were determined by using the Bio-Rad protein quantification kit. Immunoreactive proteins were detected by using enhanced chemiluminescence (ECL; GE Healthcare, Piscataway, NJ) or an Odyssey Infrared Imaging System (LI-COR Biotech, Lincoln, NE). Antibodies used for Western blotting were as follows: anti-HA, anti-Rab5, and anti-Rac1 antibodies (BD Transduction Laboratories, Lexington, KY); anti-p67phox, anti-TRAF2, anti-TRAF6, anti-IKK, anti-TRADD, anti-TNFR1, anti-IL-1R1, and anti-glutathione S-transferase (anti-GST) antibodies (Santa Cruz Biotech); and anti-Cu/ZnSOD antibody (Binding Site, Inc., San Diego, CA).
For immunoprecipitations, cells were washed with ice-cold PBS and lysed in radioimmunoprecipitation assay (RIPA) buffer at 4°C for 30min. A 500-μg aliquot of cellular protein and 5μl of primary antibody were mixed with 1ml of RIPA buffer at 4°C for 1h. A 50-μl volume of protein A-agarose beads (Santa Cruz Biotech) was then added to the mixture and rotated for 4h. The beads were washed with ice-cold PBS before experimental analyses.
In vitro kinase assays were performed with immunoprecipitated IKK by using GST-IκBα as a substrate. Kinase reactions were performed with 1μg GST-IκBα, 0.3mM cold ATP, and 10 μCi [γ-32P]ATP in 10μl kinase buffer (40mM HEPES, 1mM β-glycerophosphate, 1mM nitrophenolphosphate, 1mM Na3VO4, 10mM MgCl2, and 2mM dithiothreitol). The reaction mixtures were then incubated at 30°C for 30min. Reactions were terminated by the addition of sodium dodecylsulfate–polyacrylamide gel electrophoresis (SDS-PAGE) protein-loading buffer and heated at 98°C for 5min. After SDS-PAGE, gels were transferred to nitrocellulose membranes and exposed to x-ray film before probing with an anti-GST antibody.
In vivo localization of O2•− within endosomes was performed by using OxyBURST Green dihydro-2',4,5,6,7,7'-hexafluorofluorescein (H2HFF)-BSA (Molecular Probes, Carlsbad, CA). Stock solutions (1mg/ml) were generated immediately before use by dissolving H2HFF-BSA in PBS under nitrogen and protected from light. Cells were incubated in the presence of 50μg/ml OxyBurst Green H2HFF-BSA for 2min at 37°C and then stimulated by the addition of TNF-α (0.5ng/ml). Cells were washed with PBS and fixed in 4% paraformaldehyde for 10min. After the fixation, cells were mounted in 4',6'-diamidino-2-phenylindole (DAPI) containing antifadent and examined with fluorescent microscopy.
Nox2gp91phox-knockout mice (48) were obtained from Jackson Laboratories (strain name: B6.129S6-Cybbtm1Din/J; stock number: 002365) and were inbred on the C57BL/6 background. Primary dermal fibroblasts were generated from 1-day-old pups from control C57BL/6 or Nox2gp91phox-knockout mice, as previously described (42).
MCF-7 cells were incubated with Biotin-Transferrin (5μg/ml) at 37°C for 15min. Cells were then removed from the plates with trypsin treatment for 5min, and then carefully washed with cold PBS 3 times. Cell pellets were used to generate PNS for endosomal isolations on Iodixanol gradients, as described earlier. An equal percentage of the PNS and combined Iodixanol endosomal fractions (two through four) were loaded onto an SDS-PAGE for Western blotting with a streptavidin-conjugated 800-nm infrared dye (LI-COR Biotech). Three independent samples were evaluated in this manner, and percentage loss of endosomes during isolation was calculated [100%−(total biotin in the endosomal fraction/total biotin in the PNS)] after quantitative imaging of the Western blots on an Odyssey Infrared Imaging System from LI-COR.
Electron paramagnetic resonance spectroscopy (EPR) was used to estimate the capacity of isolated redox-active endosomes to produce NADPH-dependent O2•−. This method was previously used to specifically detect production of O2•− by isolated endosomes, which is sensitive to SOD1 but not catalase addition (32). Endosomal fractions were isolated from control (unstimulated) and TNF-α–stimulated MCF7 cells, as described earlier. The isolated endosomal fractions were mixed with 100μM diethylenetriaminepentaacetic acid (DTPA) and 50mM 5,5-dimethyl-1-pyrroline N-oxide (DMPO), in a total volume of 475μl. The reaction was initiated by adding 25μl NADPH to a final concentration of 100μM and then incubated at 37°C for 10min. Following this incubation period, each sample was immediately placed into the EPR spectrometer and scanned for ~5min at room temperature. EPR was performed using a Bruker EPR spectrometer in The University of Iowa ESR Core Facility. EPR parameters were as follows: frequency, 9.78GHz; center field, 3480G; sweep rate, 80G/21 s; power, 20.3mW; receiver gain, 5.02×104; modulation frequency, 100kHz; time constant, 81.92ms; modulation amplitude, 1.0G, sweep time, 20.972s, number of scans per spectrum, 15, and resolution, 1,024 points. EPR spectra were quantitated by peak-to-peak height of the second (low field) DMPO-OH line. Standard curves for DMPO-OH quantitation were produced by double integration of spectra with high signal-to-noise ratios [produced by using DMPO incubated with xanthine and xanthine oxidase (X/XO), data not shown]. Spectral peak areas from the X/XO standards were compared with areas of double-integrated spectra obtained (using identical EPR parameters) from standard solutions of 3-carboxy proxyl, as described previously (60). Using this method, a standard curve of peak height vs. nanomolar DMPO-OH was generated. Although these short scan times for experimental samples gave rise to lower signal-to-noise ratios than previously observed for isolated redoxosomes produced by longer scan times that average a greater number of spectra (28), the shorter scan times were necessary to minimize potential loss of the DMPO-OH adduct due to unknown enzymatic activity present in the samples. Additionally, since these EPR samples were run at room temperature, minimizing the scan time allowed a more accurately timed measurement after the 10-min 37°C incubation period.
Previous studies have implicated cellular ROS as important for TNF-α–mediated induction of NF-κB. Using a human mammary epithelial cell line (MCF-7 cells) that expresses undetectable levels of GPx-1 enzymatic activity (a cytosolic protein that degrades H2O2) (33), we first evaluated whether ectopic expression of GPx-1 could modulate transcriptional activation of NF-κB activation after TNF-α stimulation. Ectopic expression of GPx-1 from a recombinant adenoviral vector suppressed (~50%) transcriptional activation of NF-κB after TNF-α stimulation (Fig. 1A). These findings suggested that cytoplasmic H2O2 was important for NF-κB activation by TNF-α.
Using an alternative approach, we assessed whether the addition of antioxidant enzymes to the media at the time of TNF-α induction would also inhibit NF-κB activation. In these studies, the addition of 1mg/ml purified SOD1 and/or catalase to the media was used to determine the contribution of both O2•− and H2O2, respectively, in the extracellular and endosomal compartments. The addition of both SOD1 and catalase together gave maximal inhibition (~50%) of NF-κB transcriptional activation by TNF-α (Fig. 1B). The addition of SOD1 alone gave no detectable inhibition, whereas catalase addition moderately inhibited NF-κB activation by TNF-α (although not to a statistically significant level). These findings suggested that both O2•− and H2O2 must be removed to effectively inhibit TNF-α activation of NF-κB. Importantly, these redox-dependent changes in NF-κB activation were mirrored by an even greater inhibition in IKK activation (Fig. 1B, lower panel); kinase activity of IKK was significantly blocked by the addition of both SOD1 and catalase to the media at the time of stimulation with TNF-α. Cumulatively, findings in Fig. 1A and B suggest that NF-κB activation after TNF-α stimulation likely has both redox-dependent and redox-independent pathways of activation and that the IKK→NF-κB component of these pathways appears to be most redox sensitive. These findings are similar to those observed in MCF-7 cells after IL-1β stimulation (32). Because this previous study implicated NADPH-dependent ROS production by the endosomal compartment as important for IL-1β–mediated activation of IKK and NF-κB, we next investigated whether endosomal ROS production might also direct TNF-α → NF-κB signaling.
Using Iodixanol density gradient subcellular fractionation, we investigated whether TNF-α induced NADPH-dependent generation of O2•− in the endosomal compartment of MCF-7 cells. This approach was previously used with MCF-7 cells to separate endosomes from other intracellular organelles also known to produce ROS (i.e., mitochondria and peroxisomes) (32). Using lucigenin-based chemiluminescence assays, enriched vesicular fractions were assessed for their ability to generate NADPH-dependent O2•− under control (untreated) or TNF-α–stimulated conditions. As shown in Fig. 2A, TNF-α stimulation led to a substantial increase in NADPH-dependent production of O2•− in the endosomal fractions (Fr #2-4). The specificity of this production of O2•− was confirmed by using a membrane-permeable SOD mimic, MnTBAP. In the presence of MnTBAP, a significant reduction in NADPH-dependent lucigenin chemiluminescence was observed with TNF-α–stimulated endosomes (Fig. 2B). Furthermore, enhanced ROS production in the endosomal compartment after TNF-α stimulation was confirmed by endosomal loading with a membrane-impermeable BSA-conjugated O2•−-sensitive fluorescent dye, H2HFF-BSA (Fig. 2C). H2HFF-BSA was previously shown to detect SOD-sensitive ROS (presumably O2•−) in the endosomal compartment after IL-1β stimulation (32). Cumulatively, these results provide strong evidence that TNF-α activates the formation of Nox-active O2•−-producing endosomes.
Having demonstrated that TNF-α–mediated activation of NF-κB likely involves redox activation of the endosomal compartment, we next sought to determine what aspect of TNFR1 activation might be influenced by endosomal ROS. After binding of TNF-α to TNFR1, signaling is initiated by the ordered recruitments of a number of effectors and adaptors, including TRADD, TRAF2, and RIP (36, 53). TRADD is the first effector to bind to the TNF/TNFR1 complex, followed by TRAF2 and then RIP. This series of events ultimately leads to the recruitment of IKK kinases, activation of the IKK complex, and subsequently, NF-κB activation (30). We hypothesized that endosomal ROS may facilitate docking of TNFR1 effectors on the endosome that are required for IKK activation, in a similar manner to that observed for activation of the IL-1 receptor (32). By using endosomal loading with SOD1 and catalase, we evaluated the redox dependence of the recruitment of TNFR1, TRADD, and TRAF2 to the endosomal compartment after TNF-α stimulation. Peak vesicular fractions isolated from Iodixanol gradients demonstrated substantially less TRAF2 recruitment to the endosomal compartment after TNF-α stimulation in the presence of SOD1/Catalase (Fig. 2D). This reduction closely mirrored that observed in total cellular IKK activity after SOD1/catalase loading (Fig. 1B). In contrast, endosomal loading of ROS-clearance enzymes did not alter the recruitments of TNFR1 or TRADD to the endosomal compartment after TNF-α stimulation (Fig. 2D). Additionally, biochemical studies evaluating pronase and Triton-X-100 sensitivity of ROS-clearance enzymes in isolated vesicles confirmed that the enzymes were indeed loaded into the interior of the endosomes (i.e., insensitive to pronase in the absence of Triton-X-100). These findings suggested that the redox-dependent recruitment of TRAF2 to TNFR1 could be a critical step in endosomal TNF-α signaling.
The ability of endosomally loaded SOD1 and catalase to inhibit TRAF2 recruitment to TNF-α–activated endosomes, as well as NF-κB activation (Figs. 1B and and2D),2D), suggested that the formation of a redox-activated endosomal compartment may be critical for redox-dependent activation of the TNF receptor. Therefore, we next sought to investigate whether endocytosis was formally required for TNF-α–mediated redox-activation of NF-κB. To test this hypothesis, we used a dominant-negative dynamin mutant (K44A) to inhibit endocytosis. Overexpression of this dominant-negative dynamin (K44A) mutant has been shown to inhibit ~75% of transferrin-mediated uptake in MCF-7 cells (32). Furthermore, the TNF receptor also appears to depend on dynamin for uptake after ligand binding (14). Expression of the dynamin mutant inhibited TNF-α induction of NF-κB by ~40% (Fig. 3A). Furthermore, expression of dominant-negative dynamin also reduced NADPH-dependent O2•− production in isolated vesicular fractions by ~66% (Fig. 3B). Because dynamin (K44A) is ~75% effective in inhibiting endocytosis (32), these findings suggest that the majority of endosomal ROS induction after TNF-α stimulation of MCF-7 cells is likely dependent on dynamin-mediated endocytosis. In contrast, only a fraction of NF-κB activation appears to be dependent on dynamin endocytosis. We hypothesize that the dynamin-dependent fraction of NF-κB activation is redox mediated at the level of the endosome through TRAF2 recruitment and subsequent IKK activation. Such a hypothesis is consistent with the ability of SOD1 and catalase endosomal loading to more significantly inhibit IKK activation and TRAF2 recruitment, as compared with transcriptional activation of NF-κB.
Critical to understanding endosomal mechanisms for the redox activation of the TNF receptor is the identification of the source of O2•− -production. Because isolated endosomes could be induced to generate O2•− by the addition of NADPH, Nox proteins were obvious candidates. To approach this question, we first tested the sensitivity of TNF-α–induced endosomal O2•− production to diphenyleneiodonium (DPI; an NADPH oxidase inhibitor) and rotenone (a mitochondrial electron-transport chain complex I inhibitor). These studies demonstrated that O2•− production in the TNF-α–induced peak vesicular fraction was sensitive only to DPI, but not to rotenone (Fig. 4A). Our previous studies evaluating redox regulation of IL-1β signaling in MCF-7 cells revealed that Nox2 was primarily responsible for endosomal production of O2•− and redox-dependent activation of the IL-1 receptor (32). Therefore, we hypothesized that Nox2 might also be the source of endosomal production of O2•− after TNF-α stimulation. To test this hypothesis, we used Nox2 siRNA, which was previously shown to significantly reduce Nox2 protein expression in MCF-7 cells (32). As shown in Fig. 4B, Nox2 siRNA, but not scrambled siRNA, attenuated TNF-α–induced endosomal NADPH-dependent production of O2•− in the peak vesicular fractions. Because inhibition was only partially effective with Nox2 siRNA (~40%), we also tested this hypothesis in primary dermal fibroblasts from Nox2-deficient mice (KO) and their wild-type (WT) littermates. Stimulation of WT fibroblasts with TNF-α demonstrated a similar profile of NADPH-dependent production of O2•− in subcellular fractions, as seen in MCF-7 cells, with the majority of O2•− seen in peak vesicular fractions (#2-4) enriched in endosomes (Fig. 4C and D). In contrast to WT fibroblasts, Nox2 KO fibroblasts demonstrated a lack of induction in production of O2•− in the peak vesicular fraction after TNF-α stimulation (Fig. 4C and D). Furthermore, in Nox2-KO fibroblasts, TNF-α–stimulated NF-κB activation also was significantly reduced (Fig. 4D). Cumulatively, these findings strongly suggest that Nox2 is responsible for the induction of endosomal O2•− after TNF-α stimulation and that this ROS production influences NF-κB activation by this ligand.
Our current findings suggest that the recruitment of TRADD to TNF-α–activated endosomes is not dependent on the endosomal redox state (Fig. 2D). In contrast, these studies also demonstrated that TRAF2 recruitment to TNF-α–activated endosomes is dependent on endosomal ROS (Fig. 2D). We therefore hypothesized that Nox2-derived ROS were necessary for the recruitment of TRAF2 to the endosomal compartment after TNF-α stimulation. Because loading of both SOD1 and catalase into endosomes was required for maximal inhibition of IKK after TNF-α stimulation, we also predicted that the ROS derived from Nox2 that facilitated TRAF2 recruitment to endosomes was likely H2O2 (produced by dismutation of O2•−). To investigate these hypotheses, we sought to dissect the series of events whereby TRADD and TRAF2 are recruited to TNFR1, either in the plasma membrane or the endosomal compartments after TNF-α stimulation, and the extent to which these processes were dependent on H2O2.
To evaluate the recruitment of TRADD and TRAF2 to TNFR1 in the plasma membrane, we performed experiments under conditions in which cellular endocytosis was blocked (at 4°C) or significantly inhibited by adenoviral-mediated overexpression of dominant-negative dynamin. Results from these experiments (Fig. 5A) demonstrated that inhibition of endocytosis significantly impaired recruitment of TRAF2, but not TRADD, to immunoprecipitated ligand-activated TNFR1. In the absence of endocytosis at 4°C, TRAF2 recruitment to TNFR1 after TNF-α stimulation was significantly lower than that seen at 37°C. In contrast, TRADD binding to TNFR1 at 4°C was similar to that seen at 37°C in the presence of TNF-α. These findings suggested that TRADD likely recruits to TNFR1 before endocytosis, whereas recruitment of TRAF2 to TNFR1 occurs after endocytosis. This hypothesis was supported by the finding that dominant-negative dynamin expression inhibited TRAF2, but not TRADD, recruitment to TNFR1 after TNF-α stimulation (Fig. 5A).
To determine whether H2O2 facilitated the redox-dependent recruitment of TRAF2 to TNFR1, we tested whether TNF-α–dependent recruitment of TRAF2→TNFR1 could be reconstituted at the plasma membrane in the absence of endocytosis (i.e., at 4°C) by the addition of exogenous H2O2. The addition of 500μM H2O2 effectively enhanced recruitment of TRAF2 to only ligand-activated TNFR1 at the plasma membrane at 4°C (Fig. 5B). However, the level of TRAF2→TNFR1 recruitment was still significantly less than that seen at 37°C (Fig. 5B), suggesting that other enzymatic processes reduced at 4°C also may be important for the recruitment event. These findings provided a physiologic framework for TNF-α–mediated Nox2 activation in the endosomal compartment and suggest that Nox2 may be the source of H2O2 required for TRAF2 recruitment to endosomal TNFR1.
Our findings that Nox2 controls H2O2-dependent TRAF2 recruitment to endosomal TNFR1/TRADD complexes are remarkably similar to those recently described for redox activation of the IL-1β receptor (32). In the context of IL-1β, Nox2 controls H2O2-dependent TRAF6 recruitment to endosomal IL-1R1/MyD88 complexes. Hence, endocytosis of IL-1R1 and TNFR1 into a redox-active endosomal compartment appears to spatially control the H2O2-dependent recruitment of TRAFs to their cognate ligand-activated receptor complexes. We hypothesized that these redox-active signaling endosomes would share similar redox-modulator proteins, while also retaining specific factors required for the activation of their receptors (i.e., TNFR1 and IL-1R1). To formally demonstrate such a hypothesis holds, we sought to directly evaluate whether specific ligand signals (TNF-α and IL-1β) direct their cognate receptor to biologically similar Nox-active endosomes, while retaining specificity for redox-dependent recruitment of receptor-specific TRAFs.
To directly evaluate whether redox-active endosomes harbor intracellular pools of activated ligand/receptor complexes, we attempted selectively to purify Nox2-active endosomes using immuno-affinity isolation and HA-tagged Rac1. Because Rac1 is an essential Nox2 activator, we reasoned that Rac1-bound endosomes would be enriched for Nox activity and other components potentially specific to redox-active endosomes and the ligand-activated receptor. HA-Rac1–containing endosomes were isolated from untreated, TNF-α–, or IL-1β–induced MCF-7 cells. After immunoisolation, the HA-Rac1–bound and –unbound vesicular fractions were then assessed for enrichment of HA-Rac1, p67phox, SOD1, TNFR1, TRAF2, IL-1R1, and TRAF6, in relation to the common early endosomal marker, Rab5. Both p67phox (an essential activator of Nox2) and SOD1 were recently found to recruit to IL-1β–stimulated redox-active endosomes (32, 42).
Findings from these studies demonstrated that HA-Rac1 incorporation into crude vesicular fractions was significantly enhanced by both IL-1β (lane 4) and TNF-α (lane 7) stimulation (Fig. 6A). Rac1 was found only at low levels in unstimulated vesicles (lane 1). These findings support the notion that Rac1 (an essential activator of Nox2) is specifically recruited to the endosomal compartment after both IL-1β and TNF-α stimulation. Immunoaffinity isolation of HA-Rac1–bound endosomes demonstrated that the purification procedure was capable of isolating ~75% of the HA-immunoreactive endosomes (lane 5 vs. 6 and 8 vs. 9). As anticipated, this fractional enrichment for HA-Rac1 in the anti-HA-bound pellet mirrored the enrichment seen in its capacity to produce O2•−. Similarly, the enrichment of SOD1 and p67phox, relative to a general endosomal marker (Rab5), was also seen. In the absence of TNF-α or IL-1β stimulation, SOD1 and p67phoxfailed to recruit to endosomal membranes (lane 1). Most important, recruitment of IL-1R1/TRAF6 or TNFR1/TRAF2 to Rac1-containing endosomes was specific to IL-1β or TNF-α stimulation, respectively (Fig. 6, lane 5 vs. 8). No IL-1R1/TRAF6 was recruited to Rac1-containing endosomes after TNF-α stimulation (lane 8). Similarly, no TNFR1/TRAF2 was recruited to IL-1β–activated, Rac1-containing endosomes (lane 5). These findings provide direct evidence for the enrichment of ligand-activated receptors and their TRAF effectors in redox-active endosomes.
Quantifying levels of O2•− is challenging, given the unstable nature of this reactive oxygen species. Nonetheless, when thinking about the mechanisms by which redox-active endosomes transmit their ROS signals, having an estimate of the flux of O2•− production at the level of the endosome would be very useful. To this end, we attempted to estimate the flux of O2•− generated by TNF-α–stimulated redox-active endosomes. To simplify the discussion on this topic, we use the terminology set forth in a Forum Review of this issue of ARS (46), which classifies signaling redox-active endosomes as redoxosomes. We present data here that utilize EPR with DMPO standard curves to estimate the flux of O2•− produced by isolated populations of endosomes containing redoxosomes. Both TNF-α–stimulated and –unstimulated endosomes were evaluated and the difference in NADPH-dependent O2•− was designated as derived from newly formed redoxosomes. These calculations require several parameters and assumptions to calculate the flux of O2•− including (a) number of cells in the starting material, (b) number of redoxosomes per cell, (c) loss of endosomes during the isolation period, and (d) the diameter and volume of redoxosomes. The values used for each of these parameters and assumptions are given in Table 1.
EPR was used to estimate O2•− production by isolated TNF-α–stimulated redoxosomes. MCF-7 cells, treated with or without TNF-α (0.5ng/ml), were harvested at 20min after stimulation and endosomes were isolated on Iodixanol gradients. To estimate the loss of endosomes during the isolation procedure, MCF-7 cells were loaded with biotin-transferrin, and an equal percentage of the starting material for both PNS and Iodixanol endosomal fractions was evaluated by Western blot for biotin-transferrin content and quantified (Fig. 7A). Using this approach, it was calculated that 63% of the endosomal fraction was lost during the preparation procedure (Fig. 7A). Hence, a yield of 37% was used to correct for the final values of O2•− flux. Endosomes isolated from 1.0×106 MCF-7 cells were incubated with DMPO (in the presence and absence of 100μM NADPH) at 37°C for 10min and then examined by EPR (Fig. 7B). The peak height of the EPR signal was then correlated with standard curves of DMPO-OH to calculate the concentration of O2•− in each sample (Fig. 7C). As shown in Fig. 7C, control unstimulated endosomes gave a background level of production of O2•− in the presence of NADPH. This level was not significantly different from that of TNF-α–stimulated endosomes not treated with NADPH. However, O2•−-production by TNF-α–induced endosomes treated with NADPH increased threefold in comparison to the two background controls (Fig. 7C). The difference in values for NADPH-treated control vs. NADPH-treated TNF-α–induced endosomes was used to calculate the moles of O2•− generated by redoxosomes derived from 1.0×106 cells. This value was then corrected for the loss of endosomes in the isolation (63%). Based on the estimated number of redoxosomes per cell and the volume of a redoxosome (Table 1), estimates of the NADPH-dependent capacity to produce O2•− per redoxosome and per cell could be made (Table 2). These calculations took into account the 10-min reaction period and the assumption that the rate of production of O2•− was constant over that time. With these calculations, the rate of TNF-α–stimulated redoxosomal production of O2•− was estimated to be 4.3×10−22 moles redoxosome−1s−1, which is equal to ~260 molecules of O2•− redoxosome−1s−1. The errors in these calculations were±1.2×10−22 moles redoxosome−1s−1 and±74 O2•− molecules redoxosome−1s−1. By using an estimated diameter of 200nm for an early endosome/redoxosome, which has a volume of 4.2×10−18 L, one can then calculate the flux in terms of molar concentration of O2•− produced per second for a redoxosome (Ms−1). The value for the flux of O2•− in a single 200-nm-diameter TNF-α–stimulated redoxosome was estimated to be 1.0×10−4 Ms−1 at pH 7.4, with an error of±0.3×10−4 Ms−1.
For these calculations of flux, it is important also to consider the efficiency by which DMPO-OH detects O2•−. It was previously reported that this efficiency may be as low at 27% in a “clean” system (55). It is very difficult to estimate the efficiency with which DMPO-OH reports the production of superoxide in a complex biologic system since the efficiency of DMPO-OOH conversion to DMPO-OH is greatly enhanced by biologic molecules such as peroxidases. Unlike DMPO-OOH, DMPO-OH is relatively stable, so the rate of conversion of DMPO-OOH to DMPO-OH would influence the estimates of redoxosomal flux of O2•− in our studies, and biologic molecules in the sample might make this process relatively efficient. However, given previous studies on the efficiency of DMPO to detect O2•− in a clean system (55), redoxosomal O2•− flux may actually be three- to fourfold higher than estimated in our experiments. Another variable to consider is changes in redoxosome size; our estimates for the flux of O2•− are based on a diameter of 200nm (Table 1). This estimate is consistent with the literature for early endosomes (7, 8, 28, 43) and with our own studies evaluating redoxosomes by EM (42). However, endosome size is very dynamic and changes rapidly as endosomes traffic and fuse with other compartments. Since endosomal volume changes with the cube of the radius, the actual flux of O2•− will change rapidly as the size of an endosome changes. For example, a 750-nm redoxosome, which is closer to the size of a late endosome (16), would have a flux of O2•− 53-fold lower (1.9×10−6 Ms−1) than a 200-nm redoxosome (1.0×10−4 Ms−1) at pH 7.4. Hence, depending on the actual size of functioning redoxosomes, which is still a very open question, the NADPH-dependent O2•− flux may fall between these two values at neutral pH.
When translating these estimates of superoxide-producing capacity of redoxosomes to a biologic perspective, it is important to note that they do not take into account the rate of spontaneous dismutation of O2•−. Furthermore, in the EPR assay used above, it is assumed that the concentration of 50mM DMPO is sufficient to outcompete the spontaneous dismutation of O2•−. At pH 7.4, the observed second-order rate constant for spontaneous dismutation of O2•− is 2.5×105 Ms−1. This rate constant significantly increases at lower pH (kobs=6.2×106 Ms−1 at pH 6.0; kobs=2.1×107 Ms−1 at pH 5.0) (5), resulting in a significantly shorter half-life for O2•− at lower pH. Based on previous studies, the pH inside certain types of endosomal compartments can significantly decrease to as low as 5.0–6.0 for the late endosome and lysosome (4). To appreciate the potential effect of changing endosomal pH on the production of O2•− by redoxosomes, we estimated the steady-state concentration of O2•− inside the redoxosome ([O2•−]steady-state) by taking into account the rates of production and spontaneous dismutation at various pH values (Fig. 7D). These calculations assume that the production of O2•− is not affected by pH changes, an assumption that we recognize is likely false, but at present is not possible to evaluate. Using the formulas and rate constants in Table 3, the steady-state molar concentration of O2•− inside the redoxosome was calculated for pH values ranging from 5.0 to 8.0. These calculations were used to estimate the theoretical change in the steady-state level of O2•− inside the redoxosome as a function of pH (Fig. 7D). As seen in this figure, lumenal pH of the endosomal compartment will significantly influence the abundance of O2•−, as well as the dismutation product, hydrogen peroxide. At a pH of 7.4 for a 200-nm redoxosome, the steady-state concentration of O2•− is estimated to be 14μM, and this is predicted to decrease to 1.5μM at a pH of 5.0. When one superimposes potential changes in redoxosomal size (from 100 to 500nm) with potential changes in lumenal pH, the steady-state concentration of O2•− within redoxosomes could dynamically change more than two orders of magnitude (Table 3). For example, a 500-nm redoxosome at pH 5.0 would be estimated to have a steady-state lumenal concentration of O2•− of 0.40μM, whereas for a 100-nm redoxosome at pH 7.4 the value would be 41μM.
These studies assessing the redox dependence of TNF-α–mediated NF-κB activation demonstrate that endocytosis of TNFR1 and the production of endosomal ROS facilitates TRAF2 recruitment to the TNFR1/TRADD receptor complex to activate IKK. This redox-dependent TNF-α pathway shares interesting similarities with those previously observed for the IL-1β pathway (32). First, both appear to be dependent on Nox2-mediated ROS production in the endosomal compartment. Second, the redox-dependent event that controls activation of both IL-1R1 and TNFR1 appears to involve recruitment of TRAFs. Third, both receptor pathways recruit SOD1 to ligand-activated redoxosomes containing Rac1 and p67phox. In this context, redoxosomes for these two independent receptor pathways appear uniquely equipped to carry out O2•−→ H2O2 production at intracellular locations where ligand-activated receptors are concentrated.
The induction of redoxosomes after TNF-α or IL-1β stimulation provides a framework for understanding how ROS can influence receptor activation. Findings from our present study have now implicated H2O2 as a molecular signal for the recruitment of TRAF2 to the TNFR1 complex at the surface of the endosome. This finding is consistent with another study demonstrating that TRAF2-deficient mouse embryonic fibroblasts (MEFs) are resistant to ROS-induced cell death when compared with their wild-type cells (54); reconstitution of TRAF2 expression in TRAF2-deficient MEFs also restored sensitivity to ROS-induced cell death. Although our present study and this previous work suggest that TRAF2 function is responsive to redox stress, the molecular mechanism whereby TRAF2 is regulated by ROS remains unclear. We hypothesize that H2O2 may modify reactive thiol groups at certain cysteines within TRAF2, allowing its docking on the TNFR1/TRADD complex. In this context, thiol-modifying agents have been shown to inhibit TRAF2-mediated activation of the SAPK/JNK pathway (44, 45). Alternatively, Nox-dependent H2O2 could potentially modify TRAF2 function through the inhibition of cellular phosphatases, as previously shown for other redox-dependent systems (49). Although the functional significance of TRAF2 phosphorylation has been debated, recent studies have suggested that TRAF2 phosphorylation is important for CD40 signaling, as well as TNF-α–induced NF-κB activation (9, 20, 34). Of interest, Thr117 in TRAF2 is phosphorylated after TNF-α stimulation, and dephosphorylation of TRAF2 by the phosphatase PP2A holoenzyme inhibits NF-κB activation (34). Given that the PP2A holoenzyme forms a complex with TRAF2 (34), it also seems plausible that endosomal induction of H2O2 could enhance TRAF2 phosphorylation and activation by inhibiting TRAF2-associated protein phosphatases, such as PP2A.
Our studies have implicated Rac1-regulated Nox2 as a source of endosomal ROS after TNF-α stimulation. However, it appears that other Rac1-regulated Nox complexes may also serve in this context. For example, in vascular smooth muscle cells, Nox1 (which is also regulated by Rac1) appears to facilitate the induction of cellular ROS important for NF-κB activation after TNF-α stimulation (41). It is presently unclear how Nox is recruited into endosomes after TNF-α stimulation; however, studies on IL-1β endosomal activation pathways implicate Rac1 in the recruitment process (32). In the context of TNF-α, a recent study demonstrated that Nox1 associates with Rac1/TRADD/RIP complexes after TNF-α stimulation to facilitate redox-dependent necrotic death in mouse fibroblasts (27). In these studies, RIP was required for the recruitment of Nox1 to TRADD complex, but Rac1 was also required for the induction of cellular ROS. Although these studies did not evaluate the direct association of Nox1 with TNFR1 or the formation of redoxosomes, the time course of ROS production in these studies is consistent with the involvement of Nox1 in an early endosomal compartment harboring TNFR1. Hence, RIP may also be a candidate for recruitment of Nox2 to redoxosomes after TNF-α stimulation.
The finding that Nox1-dependent cellular ROS after TNFα stimulation plays an important role in necrotic cell death of primary mouse embryonic fibroblasts (27) is contrasted to findings in this report demonstrating that Nox2 facilitates NF-κB activation by TNF-α in primary dermal fibroblasts. Such differences may simply be due to the cellular origin of the fibroblasts and differential expression of Nox isoforms, or they may imply a more complex redox-dependent regulation of TNFR1 that facilitates cell death or NF-κB activation. Activation of TNFR1 has been shown to form both membrane-bound and soluble complexes that differentially regulate survival through NF-κB activation and apoptosis, respectively (22, 40). Whether these two distinct complexes differentially associate with different Nox isoforms remains to be determined.
The level of ROS production stimulated in the endosomal compartment following TNF-α stimulation has important potential implications for mechanisms of redox-signaling through TNFR1. The value for EPR-estimated redoxosomal O2•− flux of 100 μMs−1 seems very high at face value. However, when one considers the small volume of an endosome and the total O2•− produced per cell, this value takes on a new perspective. For example, the overall rate of O2•− generated theoretically by all redoxosomes in a single TNF-α–activated MCF-7 cell (4.3×10−20 moles O2•− cell−1s−1, assuming 100 redoxosomes per cell) is still ~2,000-fold lower than the rate of generation of O2•− by a single phagocyte. Neutrophils generate O2•− at a rate between 5.0×10−17 and 17×10−17 moles O2•− cell−1s−1 (11, 26, 52, 61), which translates to an estimated O2•− flux of 2.5mMs−1 within a single phagosome (61). This estimated O2•− flux for a phagosome is ~25-fold higher than that for a 200-nm TNF-α–activated redoxosome (0.1mMs−1).
Based on the similar mechanisms that facilitate TNF-α– and IL-1β–induced NF-κB activation through endosomal activation of Nox, we hypothesize that redoxosomes may play important roles in the activation of redox-dependent receptors. The activation of several receptors in addition to IL-1Rs and TNFR1 has been demonstrated to have redox-signaling components. For example, ligands such as lipopolysaccharide (LPS) (24, 51), angiotensin II (18), platelet-derived growth factor (PDGF) (56), and insulin (37) all have redox-dependent components in their signaling consistent with Nox activation at the endosomal level. To this end, we propose that redoxosomes encompass a subset of signaling endosomes that use local Nox-derived ROS to facilitate receptor-signaling events.
Based on the findings from this study and others, several common redoxosomal components have been identified (Fig. 6B). These include Nox complex components (Nox1 or Nox2, Rac1, p47phox, p67phox), SOD1, and an unidentified DIDS-sensitive O2•− channel (32, 41, 42). Although it is currently unclear whether O2•− channels also exist in TNF-α–activated redoxosomes, given the similarities seen in the redox-dependent TRAF recruitment to IL-1R1 and TNFR1 in MCF-7 cells, it is likely that such a channel exists in redoxosomes activated by both TNF-α and IL-1β. The common finding of SOD1 recruitment to both TNF-α– and IL-1β–activated redoxosomes also suggests that SOD1 may play an active role in O2•− dismutation at the surface of the endosome. Such SOD1 functions may be important for mediating localized H2O2-directed signaling events and/or the redox-sensitive control of Rac1-regulated Nox complexes, as recently described (19).
This work was supported by NIDDK (RO1 DK067928 and DK051315), the vector core funded through the Center for Gene Therapy (P30 DK54759), and the Roy J. Carver Chair in Molecular Medicine. We also gratefully acknowledge the support of The University of Iowa ESR Facility and Dr. Christine Blaumueller for editorial assistance.
BSA, bovine serum albumin; DAPI, 4', 6'-diamidino-2-phenylindole; DMPO, 5,5-dimethyl-1-pyrroline N-oxide; DPI, diphenyleneiodonium; DTPA, diethylenetriaminepentaacetic acid; Duox, dual oxidase; EDTA, ethylenediaminetetraacetic acid; EGF, epidermal growth factor; FBS, fetal bovine serum; GPx1, glutathione peroxidase 1; GST, glutathione-s-transferase; HA, influenza A viral hemagglutinin (tag); HEPES, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid; (H2HFF)-BSA, dihydro-2',4,5,6,7,7'-hexafluorofluorescein-bovine serum albumin; H2O2, hydrogen peroxide; IL-1β, interleukin-1β; IL-1R1, interleukin-1 receptor 1; IKK, IκB kinase; JNK, C-Jun N-terminal kinase; LPS, lipopolysaccharide; MEF, mouse embryonic fibroblast; MOI, multiplicity of infection; MyD88, myeloid differentiation primary response gene (88); NADPH, nicotinamide–adenine dinucleotide phosphate (reduced); NF-κB, nuclear factor κB; Nox, NADPH oxidase; O2•−, superoxide; PBS, phosphate-buffered saline; PDGF, platelet-derived growth factor; PNS, postnuclear supernatant; PP2A, protein phosphatase 2A; Rac1, Ras-related C3 botulinum toxin substrate 1; RIPA, radioimmunoprecipitation assay; RIP, receptor-interacting protein; ROS, reactive oxygen species; SAPK, stress-activated protein kinase; SDS-PAGE, sodium dodecylsulfate–polyacrylamide gel electrophoresis; SOD1, superoxide dismutase 1; siRNA, small interfering RNA; TNF-α, tumor necrosis factor α; TNFR1, tumor necrosis factor receptor 1; TRAF2, tumor necrosis factor receptor-associated factor 2; TRADD, tumor necrosis factor receptor-associated death domain.
No competing financial interests exist.