Department of Biological Sciences, Purdue University, West Lafayette, Indiana, USA
Bindley Bioscience Center, Purdue University, West Lafayette, Indiana, USA
Address correspondence and reprint requests to Dr. Daniel Suter, Department of Biological Sciences, Purdue University, 915 West State Street, West Lafayette, IN 47907-2054, USA. E-mail: firstname.lastname@example.org
NADPH oxidases are important for neuronal function but detailed subcellular localization studies have not been performed. Here, we provide the first evidence for the presence of functional NADPH oxidase 2 (NOX2)-type complex in neuronal growth cones and its bidirectional relationship with the actin cytoskeleton. NADPH oxidase inhibition resulted in reduced F-actin content, retrograde F-actin flow, and neurite outgrowth. Stimulation of NADPH oxidase via protein kinase C activation increased levels of hydrogen peroxide in the growth cone periphery. The main enzymatic NADPH oxidase subunit NOX2/gp91phox localized to the growth cone plasma membrane and showed little overlap with the regulatory subunit p40phox. p40phox itself exhibited colocalization with filopodial actin bundles. Differential subcellular fractionation revealed preferential association of NOX2/gp91phox and p40phox with the membrane and the cytoskeletal fraction, respectively. When neurite growth was evoked with beads coated with the cell adhesion molecule apCAM, we observed a significant increase in colocalization of p40phox with NOX2/gp91phox at apCAM adhesion sites. Together, these findings suggest a bidirectional functional relationship between NADPH oxidase activity and the actin cytoskeleton in neuronal growth cones, which contributes to the control of neurite outgrowth.
We have previously shown that reactive oxygen species (ROS) are critical for actin organization and dynamics in neuronal growth cones as well as neurite outgrowth. Here, we report that the cytosolic subunit p40phox of the NOX2-type NADPH oxidase complex is partially associated with F-actin in neuronal growth cones, while ROS produced by this complex regulates F-actin dynamics and neurite growth. These findings provide evidence for a bidirectional relationship between NADPH oxidase activity and the actin cytoskeleton in neuronal growth cones.
Reactive oxygen species (ROS) oxidize various molecules, including proteins, lipids, and DNA. When ROS levels are elevated, this activity can lead to oxidative stress, cell death, and aging (Sohal and Orr 2012) and contribute to several chronic and degenerative diseases, including cancer, Alzheimer's, and Parkinson's disease (Hernandes and Britto 2012; Yang et al. 2013). On the other hand, a growing body of literature indicates that ROS also act as important physiological signaling molecules in cell proliferation, differentiation, motility, and apoptosis (Bedard and Krause 2007; Finkel 2011). Accordingly, ROS are not only uncontrolled by-products of aerobic metabolism but are also specifically generated by NADPH oxidases, the mitochondrial respiratory chain, and lipoxygenases (Camello-Almaraz et al. 2006; Bedard and Krause 2007; Taddei et al. 2007). A tight control of cellular ROS concentration is essential to ensure specific signaling. Perturbing this redox balance can result in the aforementioned diseases.
Because of the highly reactive and short-lived nature of ROS (Winterbourn 2008), intracellular ROS signaling likely has to occur within close vicinity of the ROS source. Therefore, localized activation seems essential for ROS signaling. In non-neuronal cells, NADPH oxidases have been localized to distinct subcellular regions involved in cell adhesion and migration, including leading edge, ruffles, and focal adhesions (Ushio-Fukai 2006). Accordingly, ROS derived from NADPH oxidases have been implicated in adhesion of fibroblasts (Chiarugi et al. 2003) and in mig-ration of endothelial cells (Moldovan et al. 2000; Ushio-Fukai et al. 2002; Ikeda et al. 2005), HeLa cells (Nimnual et al. 2003; Kim et al. 2009), smooth muscle cells (Schroder et al. 2007; Lee et al. 2009), and keratinocytes (Kim et al. 2011). Whether ROS produced by NADPH oxidase regulate adhesive and motile processes in neurons such as growth cone protrusion, neurite outgrowth, and axon guidance is not clear.
The family of NADPH oxidases consists of seven members, which all contain a major membrane-bound flavocytochrome b558 enzymatic subunit but differ with respect to the composition of additional membrane-bound and cytoplasmic subunits (Bedard and Krause 2007). The first NADPH oxidase to be characterized was found in phagocytes and contained NOX2/gp91phox (referred to as ‘NOX2’ in the remainder of this article). The fully assembled and active NOX2 complex includes the p22phox, Rac1, p47phox, p67phox, and p40phox subunits, which regulate the enzymatic activity of the NOX2 complex. NADPH oxidase family members NOX1, NOX2, NOX3, and NOX4 are expressed in different portions of the nervous system, particularly in neurons, microglia, and astrocytes (Sorce and Krause 2009; Hernandes and Britto 2012). NADPH oxidase-derived ROS have been implicated in hippocampal synaptic plasticity and memory formation (Kishida et al. 2006), NMDA receptor activation (Brennan et al. 2009), nerve growth factor-induced neuronal differentiation and neurite outgrowth of PC-12 cells (Suzukawa et al. 2000; Ibi et al. 2006), and neuronal apoptosis (Tammariello et al. 2000; Guemez-Gamboa and Moran 2009). On the other hand, microglial cells and proinflammatory cytokine-treated neurons release NADPH oxidase-derived superoxide leading to neuronal toxicity (Barth et al. 2012), as described in Alzheimer's and Parkinson's disease (Sorce and Krause 2009; Gao et al. 2012). We have recently reported that ROS derived from NADPH oxidases regulate F-actin organization, dynamics, and neurite outgrowth (Munnamalai and Suter 2009); however, the exact subcellular localization and interactions of NADPH oxidase with the actin cytoskeleton in neuronal growth cones have not been investigated.
Here, we report on the first localization of a NOX2-type NADPH oxidase in neuronal growth cones. NADPH oxidase inhibition with VAS2870 or celastrol resulted in reduced retrograde F-actin flow and neurite outgrowth, confirming our earlier results. NADPH oxidase activation with a protein kinase C (PKC) activator resulted in increased ROS levels in the growth cone periphery. We found that the regulatory cytosolic subunit p40phox exhibited F-actin association in unstimulated growth cones and little colocalization with plasma membrane-bound NOX2. However, upon growth cone stimulation with the Aplysia cell adhesion molecule (apCAM), p40phox and NOX2 accumulated and colocalized at adhesion sites. In summary, these findings point toward an interesting bidirectional relationship between NADPH oxidase and the actin cytoskeleton in neuronal growth cones.
Materials and methods
Aplysia bag cell neuronal culture
Aplysia bag cell neurons were plated on coverslips coated with 20 μg/mL poly-L-lysine (70–150 kD) as previously described (Lee et al. 2008; Suter 2011). Cells were kept in L15 medium (Invitrogen, Carlsbad, CA, USA) supplemented with artificial seawater (L15-ASW) overnight at 14°C. All reagents were from Sigma (St. Louis, MO, USA) unless otherwise specified.
Live-cell imaging and neurite outgrowth analysis
For live-cell imaging, a chamber containing Aplysia bag cell neurons cultured for 1 day was prepared as recently described (Lee et al. 2008; Suter 2011). 0.1% dimethylsulfoxide (DMSO) (vehicle control), 5 μM VAS2870 (Enzo Life Sciences, Farmingdale, NY, USA) or 5 μM celastrol (Cayman Chemical, Ann Arbor, MI, USA) dissolved in L15-ASW were sequentially applied to live neurons, while cells were observed with a 60× 1.4 NA oil objective (plus 1.5× magnification) using a Nikon TE2000 Eclipse microscope (Nikon, Melville, NY, USA; Munnamalai and Suter 2009). An OG590 long-pass red filter (Chroma Technology Corp., Bellows Falls, VT, USA) was used to eliminate shorter wavelengths harmful in live-cell imaging. Each drug was pre-incubated for 5 min before differential interference contrast images were recorded at 10-s intervals. VAS2870 and celastrol were applied for a total of 20 min. The drug was then washed out with fresh L15-ASW. For retrograde flow analysis, three kymographs of differential interference contrast time-lapse sequences were constructed from each growth cone using line scans across the peripheral (P) domain (Fig. 1; kymograph function in MetaMorph 7 software; Molecular Devices, Downingtown, PA, USA). Retrograde flow rates of individual growth cones were determined by averaging five velocity measurements from each kymograph.
Neurite outgrowth experiments were carried out similarly as previously reported (Munnamalai and Suter 2009). Neurons were imaged in L15-ASW medium 4 h after plating using a 10× phase objective. The medium was then replaced with either L15-ASW, medium containing 0.05% DMSO or 2.5 μM VAS2870 or 1 μM celastrol. Cells were incubated at 14°C for 24 h prior to a second round of imaging. Drugs were then washed out extensively with fresh L15-ASW and cells were incubated at 14°C for an additional 24 h prior to the collection of a final set of images. Total neurite length was calculated by summing the lengths of all neurites per cell. Only cells with total neurite length of x ≥ 50 μm were included in the analysis.
NADPH oxidase immunostaining and fluorescence microscopy
After 1 day in culture, bag cell neurons were fixed with 3.7% formaldehyde/400 mM sucrose/ASW, pH 7.6 for 20 min at 20°C. Following fixation, cells were permeabilized with 0.1% saponin in fixation solution for 15 min. After three washes with phosphate-buffered saline (PBS)/0.01% saponin, cells were incubated with Alexa 488-phalloidin (1 : 100; Invitrogen) in PBS/0.01% saponin for 20 min. Following three washes, cells were blocked with 5% bovine serum albumin in PBS/0.01% saponin for 30 min. For detection of NOX2, the rabbit anti-gp91phox antibody R2085 was incubated at 1 : 500 in blocking solution for 1 h. This antibody was raised against a peptide corresponding to AA 548-560 in human gp91phox (Quinn et al. 1989). For peptide blocking experiments (Figure S2), we used the human NOX2 548-560 peptide KQSISNSESGPRG or the corresponding Aplysia NOX2 peptide KHCNRFSSSESK (XP_005090645). Therefore, the NOX2 antibody was diluted at 1 : 500 in blocking solution together with the same dilution of either the human or Aplysia peptide (10 mg/mL stock; Genscript, Piscataway, NJ, USA) over night at 4°C before the antibody was applied to fixed neurons. A mouse anti-p40phox was used at a 1 : 200 in blocking solution for 1 h at 20°C. This antibody was raised against full length human p40phox (Gauss et al. 2002). apCAM was detected with the monoclonal antibody 4E8 diluted to 10 μg/mL in blocking solution (Suter et al. 1998). Following three washes, primary antibodies were detected by applying either 4 μg/mL Alexa 568 or Alexa 647 labeled goat-anti-rabbit or mouse secondary antibodies (Invitrogen) in the same blocking solution for 30 min at 20°C. Fluorescent images were acquired in 20 mM n-propyl-gallate in PBS/80% glycerol pH 8.5 on a Nikon TE2000 Eclipse microscope equipped with a 60X 1.4 NA oil (plus 1.5× magnification) objective, an X-cite 120 metal halide lamp (EXFO, Mississauga, ON, Canada), appropriate fluorescence filter sets (Chroma, Bellows Falls, VT, USA), Cascade II cooled CCD camera (Photometrics, Tucson, AZ, USA), and MetaMorph 7 software. Neuronal cell bodies were imaged by acquiring z-stacks at 10 -μm intervals on a Zeiss LSM 710 confocal microscope (Carl Zeiss Microscopy GmbH, Jena, Germany) equipped with 488, 568, and 610 nm lasers and appropriate filters.
For quantitative ROS imaging, we used a hydrogen peroxide (H2O2)-specific fluorescent dye Peroxyfluor-6 acetoxymethyl ester Peroxyfluor-6 (PF-6) kindly provided to us by Dr. Christopher Chang, University of California, Berkeley, CA, USA (Dickinson et al. 2011). After 1 day in culture, growth cones were incubated for 20 min in the dark with ASW containing 5 μM PF-6 (for H2O2 detection) together with 250 nM Calcein Red-Orange AM (Invitrogen, Life Technologies, Grand Island, NY, USA), a cell permeable fluorescent dye used as volume marker for ratiometric imaging (Li et al. 2009). Three washes with ASW were performed to remove free dye. Pre-treatment images (time 0) were captured using a Nikon TE2000 Eclipse microscope using a 480-nm excitation and 535-nm emission filter for PF-6 and a 575-nm excitation and 610-nm emission filter for Calcein Red-Orange. The growth cones were then treated with ASW containing 0.1% DMSO, 0.1 μM phorbol 12,13-dibutyrate (PDBu; a protein kinase C activator), 5 μM VAS2870 or a com-bination of PDBu and VAS2870 for up to 15 min before a second set of images were taken of the same growth cones. In one set of experiments, the second image was taken 15 min after addition of VAS2870 (Fig. 2d–h). In another set, imaging was done every 30 s after addition of the drugs (Fig. 2i).
Average fluorescent intensities of both PF-6 and Calcein Red-Orange signals were measured in three regions in the growth cone periphery and then background subtracted using MetaMorph software. Each PF-6 and Calcein Red-Orange measurement was then normalized by the average DMSO control value at the same time point to correct for experimental variability. The DMSO-normalized signals were then volume corrected by dividing PF-6 values by the corresponding Calcein Red-Orange values. Volume-corrected PF-6 signals were calculated as% changes of H2O2 levels at specific time points by dividing each value by the time 0 value (Fig. 2i). PF-6 signals in DMSO control changed relatively little within the first 4 min (~2–5% reduction; Fig. 2i), while more significantly after 15 min of DMSO treatment (~15% reduction; Fig. 2h).
To disrupt F-actin bundles only, growth cones were treated with 350 nM cytochalasin B for 5 min and then fixed and stained for F-actin, NOX2, and p40phox (Burnette et al. 2008). To strongly inhibit actin assembly in lamellipodia and filopodia, growth cones were treated with 5 μM cytochalasin B for 1 min prior to fixation and staining for F-actin, NOX2, and p40phox (Forscher and Smith 1988).
Hypotonic buffer (10 mM Tris HCl pH 7.5, 10 mM NaCl, 3 mM MgCl2, 1 mM EDTA, 1 mM EGTA) was supplemented with protease inhibitors (0.5 mM Pefabloc SC Plus (Roche Diagnostics Corporation, Indianapolis, IN, USA); 1% Protease Inhibitor Cocktail), phosphatase inhibitors (10 mM NaF, 1 mM Na3VO4), and cytoskeletal stabilizing agents (20 μM taxol and 5 μM phalloidin). Total CNS tissue from two Aplysia was collected in 4 mL of ice-cold hypotonic buffer, cut into smaller pieces with dissecting scissors, and homogenized with an Omnimixer. All steps were performed at 4°C. The lysate was further homogenized with a Dounce tissue homogenizer (10 strokes) and then cleared by centrifugation at 3000 g for 10 min to remove nuclei and cell debris. The remaining supernatant (S1) was spun at 100 000 g in a tabletop ultracentrifuge (TLA 100.3 rotor; Beckman Coulter, Inc., Brea, CA, USA) for 1 h to separate membrane and cytoskeletal (P2) fraction from cytoplasmic (S2) fraction. The membrane and cytoskeletal fraction (P2) was resuspended in hypotonic buffer containing 1% Triton X-100 followed by centrifugation at 100 000 g for 1 h to separate the membrane (S3) and cytoskeletal (P3) fractions. The membrane fraction was treated with DNase 1 for 30 min to reduce the amount of membrane-associated actin cytoskeleton. The enriched membrane fraction was collected as supernatant (S4) after centrifugation at 100 000 g for 30 min. The cytoskeletal fraction (P3) was treated with 20 μM cytochalasin B for 30 min on ice followed by cen-trifugation at 100 000 g for 30 min, resulting in a supernatant (S5) containing mainly monomeric actin and a pellet (P5) containing remaining F-actin. The fractions were boiled at 70° C for 10 min, resolved on a 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis gel, and then transferred onto a polyvinylidene difluoride membrane. Following blocking with 5% non-fat dry milk in PBS, the membrane was probed with these primary antibodies: rabbit anti-actin (Sigma, A5060; 1 : 2000), rabbit anti-NOX2 (1 : 5000), and mouse anti-p40phox (1 : 1000) in PBS/0.05% Tween 20/0.1% non-fat dry milk. We used the secondary antibodies Alexa 680-goat-anti-rabbit and Alexa 800-goat-anti-mouse (Invitrogen) at 0.4 μg/mL in PBS/0.05% Tween 20/0.1% non-fat dry milk/0.01% sodium dodecyl sulfate, followed by detection with the Odyssey system (LI-COR Biosciences, Lincoln, NE, USA). For quantification, the integrated densities of the NOX2 and p40phox bands were determined for both the cytoskeletal and membrane fraction using Image J software (http://imagej.nih.gov/ij/) (Fig. 6b). The ratio of the cytoskeletal over membrane signal was determined and then averaged from three independent experiments.
Actin Coimmunoprecipitation from Aplysia CNS Lysate
Protein A-sepharose beads (#17-1279-01; GE Healthcare, Pittsburgh, PA, USA) were washed with PBS. All steps were performed at 4°C. 100 μL of washed protein A sepharose was pelleted, resuspended in 300 μL PBS containing 1 μg/μL of rabbit anti-actin antibody (Sigma, A5060), and incubated on a rotator for 1 h at 4°C. The antibody was cross-linked to protein A-sepharose using disuccinimidyl glutarate (#20593, DSG; Pierce, Thermo Fisher Scientific, Rockford, IL, USA), prepared immediately before use. DSG dissolved in DMSO was added to a final concentration of 5 mM and incubated on a rotator for 2 h. DSG was then removed by pelleting the mixture at 4000 g for 1 min. The cross-linked beads were washed twice with PBS, twice with 0.1 M glycine HCl pH 2.5 to elute any unbound antibody, and three times with PBS before beads were stored in PBS at 4°C or used for immunoprecipitation. Total CNS tissue from two Aplysia was collected in 4 mL of complete ice-cold lysis buffer [50 mM Tris HCl pH 7.5, 150 mM NaCl, 1 mM EGTA; 1% protease inhibitor cocktail (Cytoskeleton, Inc, Denver, CO, USA)], cut into smaller pieces with dissecting scissors, incubated in lysis buffer for 10 min on ice, and homogenized with an Omnimixer. The homogenized lysate was spun at 12 000 g for 20 min at 4°C, and the remaining supernatant was used for the immunoprecipitation. 100 μL of fresh lysate was added to 20 μL of beads with cross-linked anti-actin antibody and incubated on a rotator over night at 4°C. Following centrifugation at 4000 g for 1 min, the beads were washed five times with PBS and bound proteins were eluted by resuspending the beads in 75 μL PBS and 25 μL gel loading buffer and boiling at 100°C for 10 min. The boiled samples were spun as before and eluted proteins were analyzed by western blotting for both actin and p40phox.
The restrained bead interaction assay (RBI) with apCAM-coated beads was performed as described (Suter et al. 1998, 2004). Beads were placed on the P domain of the growth cone and restrained with a glass micropipette until central (C) domain extension and protrusion of the leading edge occurred. When the C domain reached the bead, the cells were fixed and stained with NOX2 and p40phox antibodies.
Image processing, fluorescence intensity, and colocalization analysis
Photoshop CS3 extended (Adobe, San Jose, CA, USA) was used for image processing and making videos. Canvas X (ACD Systems of America, Inc, Miami, FL, USA) was used for final figure assembly. For quantification of NOX2 and p40phox signals, the average NOX2 and p40phox fluorescence intensities per unit area was determined in the growth cone P domain using MetaMorph 7.0, followed by background subtraction and normalization to control values in case of cytochalasin treatments. For 5 μM cytochalasin B treatments, background-subtracted fluorescence intensities of NOX2 and p40phox were measured in the P domain region that has been cleared from F-actin. For quantification of p40phox colocalization with NOX2 or F-actin, each signal was thresholded at the level of the average intensity plus 10% for a given area. After thresholding, the colocalization analysis function in MetaMorph was used to determine the% area overlap of p40phox with either NOX2 or F-actin or the% area overlap of NOX2 with either p40phox or F-actin.
Quantified data are presented as mean values ± SEM. STATISTICA 10 Academic software (StatSoft, Inc., Tulsa, OK, USA) was used for statistical data analysis. Significant differences between two conditions were determined with Student's t-tests if data showed normal distribution. Comparison of multiple conditions was performed with a one-way anova followed by Dunnett's T3 post hoc test. Significance levels were labeled the following way: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.
NADPH oxidase inhibition negatively impacts the actin cytoskeleton and neurite outgrowth
We previously reported that lowering cytosolic ROS levels in Aplysia neuronal growth cones by either a general radical scavenger, α-N-tert-butyl-phenylnitrone, or by NADPH oxidase inhibition via phenylarsine oxide and apocynin resulted in reduced actin content, impaired actin dynamics, and decreased neurite outgrowth (Munnamalai and Suter 2009). Phenylarsine oxide is a potent NOX2 inhibitor, but being a vicinal dithiol-binding reagent it is not exclusively specific to NADPH oxidase (Le Cabec and Maridonneau-Parini 1995; Jaquet et al. 2009). Apocynin is still widely used as NADPH oxidase inhibitor; however, there is now evidence that it may act as antioxidant and require the presence of myeloperoxidase to function as a NADPH oxidase inhibitor (Heumuller et al. 2008; Jaquet et al. 2009; Wind et al. 2010). To confirm that our previous results were indeed caused by reduced NADPH oxidase activity, we repeated oxidase inhibition experiments with one of the most specific chemical pan-NADPH oxidase inhibitors currently available, VAS2870 (ten Freyhaus et al. 2006; Wind et al. 2010; Altenhofer et al. 2012).
Treatment of Aplysia growth cones with 5 μM VAS2870 for 20 min reduced transition (T) zone ruffling, flattened the peripheral (P) domain, caused extension of the central (C) domain and filopodial lengthening when compared to DMSO control pre-treatment (Fig. 1a–c; Video S1). These effects were partially reversible after 60 min following drug washout (Fig. 1d). NADPH oxidase inhibition significantly decreased the F-actin content in T zone ruffles, as well as in filopodial actin bundles and lamellipodial actin networks in the P domain (Fig. 1e–h). Treatment with 5 μM VAS2870 reduced retrograde actin flow rates by 35% compared to DMSO pre-treatment (Fig. 1i; mean flow rates ± SEM were 4.6 ± 0.1 μm/min in DMSO vs. 3.0 ± 0.1 μm/min in VAS2870). Flow rates nearly recovered to control values following drug washout. Total neurite length per neuron was strongly reduced in neurons treated with 2.5 μM VAS2870 for a 24 h when compared to medium only or DMSO control conditions (Fig. 1j; 81% reduction compared to DMSO). Extensive drug washout over the following 24-h period did not reverse neurite length values. This is likely because of the significantly longer incubation time when compared to the results of the 20-min VAS2870 treatments. Using a different NADPH oxidase inhibitor, celastrol (Jaquet et al. 2011), we observed similar effects on growth cone morphology, actin flow, and neurite outgrowth (Figure S1). In summary, we found that treatment of cultured Aplysia neurons with VAS2870 or celastrol resulted in similar effects as our previous approaches of NADPH oxidase inhibition (Munnamalai and Suter 2009), confirming our findings that NADPH oxidase activity is critical for controlling actin organization and dynamics in growth cones as well as neurite growth.
Localization of a functional NOX2-type NADPH oxidase in neuronal growth cones
Although NOX1, NOX2, and NOX4 NADPH oxidase have been previously detected in cultured cerebellar granule neurons (Coyoy et al. 2008) and NOX2 additionally in hippocampal neurons (Tejada-Simon et al. 2005; Park and Jin 2008), sympathetic neurons (Tammariello et al. 2000; Hilburger et al. 2005), and dorsal root ganglion neurons (Cao et al. 2009), none of these studies investigated the localization of these enzyme complexes specifically in growth cones. Here, we identified a NOX2-type NADPH oxidase complex in growth cones using antibodies raised against human homologs of the membrane-bound NOX2 (Quinn et al. 1989) and cytosolic p40phox subunits (Gauss et al. 2002). Western blotting of Aplysia CNS protein extracts with these antibodies detected a major band at the expected molecular weights of 75 kDa and 40 kDa, respectively (Fig. 2a). NOX2 proteins run as broad band between 73 and 91 kDa on SDS-PAGE, and the band pattern depends on the species and glycosylation (Quinn et al. 1989; Gauss et al. 2002; Krijnen et al. 2003).
Immunolabeling of cultured Aplysia bag cell neurons with these NOX2 and p40phox antibodies revealed a punctate pattern of both proteins within growth cones (Fig. 2b–c). NOX2 labeling in the P domain is more uniform and dense than p40phox labeling, which is expected for a plasma membrane-associated protein. NOX2 signals were markedly higher in the C domain of the growth cone (Fig. 2b) compared to the P domain, which can be attributed to NOX2's association with vesicles. Pre-incubation of the rabbit anti-NOX2 antibody with either the human peptide used for raising the antibody or the corresponding Aplysia peptide significantly reduced the fluorescence intensity in the growth cone periphery (Figure S2), providing evidence for the specificity of the NOX2 immunolabeling. Double labeling of NOX2 and the cell adhesion molecule apCAM revealed a partial colocalization between the two membrane-bound proteins. While apCAM was homogenous across the growth cone, NOX2 was more concentrated in discrete puncta (Figure S3). Further evidence for plasma membrane association of NOX2 in unstimulated Aplysia neurons is given by confocal imaging of the large neuronal cell body (Figure S4). The localization of the cytosolic p40phox subunit differed from NOX2 in a few ways. First, p40phox exhibited a striking array-like localization (Fig. 2c), very similar to F-actin bundles in the P domain. Second, the labeling in the C domain was not increased as in the case of NOX2. Third, hardly any p40phox was found in very thin growth cone regions that were clearly NOX2 positive (arrowhead in Fig. 2c and 3c).
NADPH oxidase produces superoxide, which can be converted quickly into H2O2, a ROS that has a longer half-life than superoxide and is likely involved in many ROS-mediated signaling processes (Bedard and Krause 2007; Winterbourn 2008). We used Peroxyfluor-6 acetoxymethyl ester (PF-6) as a H2O2-specific fluorescent dye (Dickinson et al. 2011) in combination with Calcein Red-Orange AM as a volume marker to perform ratiometric quantification of H2O2 levels in the growth cone P domain following activation and inhibition of NADPH oxidase (Fig. 2d–i). PF-6 signals were the highest in the C domain, specifically in mitochondria, a well-known ROS source, as well as in T zone ruffles. NADPH oxidase inhibition with 5 μM VAS2870 reduced H2O2 levels more than the DMSO control treatment after 15 min (Fig. 2d–h; 30% reduction by VAS2870; 16% reduction by DMSO), although these reductions were not significantly different (Student's t-test; p = 0.12). Conversely, we used the phorbol ester phorbol 12,13-dibutyrate (PDBu) to activate PKC, which is known to stimulate activity of NADPH oxidase through phosphorylation of p47phox (Fontayne et al. 2002). Within 30 s of PDBu application, H2O2 levels increased by 12 ± 3%. This PDBu-induced rise in H2O2 levels could be significantly decreased by coincubation with 5 μM VAS2870 to levels that were similar to basal H2O2 levels (Fig. 2i). In summary, immunolocalization and ROS imaging data further confirmed the presence of a functional NADPH oxidase complex in neuronal growth cones.
p40phox localization is F-actin dependent
Since p40phox exhibited an F-actin-like distribution, we performed triple labeling with phalloidin, p40phox, and NOX2 antibodies to determine a potential colocalization between p40phox and F-actin (Fig. 3). In the growth cone P domain, p40phox exhibited only a partial overlap with NOX2 (Fig. 3e; inset e'). 38 ± 2% of p40phox colocalized with NOX2, while 22 ± 1% of NOX2 overlapped with p40phox (Fig. 3g). The highest colocalization between p40phox and NOX2 was found in the neck of the growth cone, which consolidates into the axon, as well as in contact areas between two adjacent growth cones (arrow in Fig. 3e). Little colocalization between p40phox and NOX2 was detected at the plasma membrane of neuronal cell bodies, while we found punctate structures inside the cell body that were labeled with both antibodies (Figure S4). On the other hand, p40phox exhibited higher colocalization with filopodial F-actin bundles in the growth cone periphery (Fig. 3f; inset f'), when compared to NOX2-colocalization with F-actin (Fig. 3g; 56 ± 3% vs. 47 ± 2%; *p < 0.05).
To further investigate the relationship between p40phox, NOX2, and F-actin, we analyzed their distribution in growth cones with different amounts of F-actin. Most growth cones exhibit a typical actin organization with filopodial actin bundles in the P domain (growth cone in lower left of Fig. 4). These growth cones were rich in p40phox in the P domain. A smaller fraction of flat growth cones that are devoid of F-actin had very little p40phox in the P domain (growth cone in upper right of Fig. 4). These growth cones only exhibited p40phox signals in areas where the microtubule- and organelle-rich C domain invades the F-actin-free P domain. On the other hand, NOX2 signals in the P domain did not vary that much between growth cones containing different amounts of F-actin (Fig. 4c).
To confirm this correlation between p40phox and the F-actin localization, growth cones were treated with two different concentrations of cytochalasin B to perturb the F-actin organization (Fig. 5). A low dose of cytochalasin B (350 nM) resulted in the disruption of F-actin bundles but preservation of the F-actin meshwork (Burnette et al. 2008). Along with the disruption of F-actin bundles, we observed a 32% decrease in p40phox levels within the P domain (Fig. 5j; anova, **p < 0.01). A high dose of cytochalasin B (5 μM) strongly inhibited F-actin assembly in bundles and meshwork in the P domain. As retrograde actin flow continued, this treatment resulted in clearing of all F-actin structures from of the P domain besides a cytochalasin-resistant band of F-actin at the leading edge (Forscher and Smith 1988). p40phox localization was completely disrupted in the region of actin clearance, and its levels decreased by 74% (Fig. 5i–j; ****p < 0.0001). Interestingly, NOX2 levels were also reduced by cytochalasin B treatment (17% by 350 nM cytochalasin B, anova, **p < 0.01; 58% by 5 μM cytochalasin B, ****p < 0.0001). In summary, our results suggest that the growth cone localization of p40phox, as well as of NOX2, depends on F-actin.
F-actin association of p40phox
To further demonstrate a potential association of p40phox with the F-actin cytoskeleton, we performed subcellular fractionation of Aplysia CNS tissue lysates (Fig. 6a). As expected, NOX2 was enriched in the plasma membrane compared to the cytoskeletal fraction and was not removed from this fraction following DNase 1 treatment, which reduces membrane-associated F-actin (Fig. 6a). On the other hand, p40phox was more prominent in cytoskeletal than in membrane fractions. The distribution of p40phox in the various cytoskeletal fractions (cytoskeletal, cytochalasin B-sensitive cytoskeletal, cytochalasin B-resistant cytoskeletal) showed a similar pattern as actin, indicating that a significant portion of p40phox fractionates together with F-actin. The ratio of cytoskeletal over membrane-associated protein was higher for p40phox when compared to NOX2 (Fig. 6b; 1.08 ± 0.05 for p40phox vs. 0.76 ± 0.03 for NOX2; paired Student's t-test, p = 0.05). In addition, we detected p40phox in actin immunoprecipitations from Aplysia CNS lysates (Fig. 6c). In summary, immunolocalization and biochemical data strongly suggest that the cytosolic subunit p40phox associates with the F-actin cytoskeleton in unstimulated neurons, while localization of the membrane subunit NOX2 appears less dependent on F-actin.
Enhanced p40phox-NOX2 colocalization in apCAM-evoked growth
To investigate whether p40phox and NOX2 distribution changes following growth cone stimulation with adhesive cues, we performed the restrained bead interaction (RBI) assay utilizing beads coated with the Aplysia cell adhesion molecule apCAM (Suter et al. 1998, 2004). Since considerable actin remodeling occurs around beads during apCAM-evoked growth, we tested whether NOX2 and p40phox localization was altered during these events. We found a significant increase of both p40phox and NOX2 signals in vicinity of apCAM beads when compared to adjacent non-RBI areas (Fig. 7a–d). Furthermore, there was a significantly higher colocalization of p40phox and NOX2 when compared with non-RBI areas (Fig. 7d–e; 71.4 ± 4.0% colocalization in bead areas vs. 29.5 ± 4.8% in non-RBI areas; n = 4 RBI experiments; Student's t-test: ***p < 0.001). In the example shown here, the staining for NOX2 and p40phox appeared reduced in the C domain that extended toward the restrained bead when compared to C domain labeling of control growth cones. However, we did not observe this reduction consistently in other experiments. These findings indicate that apCAM-mediated adhesion could result in NOX2 cytosolic subunit recruitment and potential NADPH oxidase activation that involves F-actin remodeling.
This is the first report on subcellular localization of NADPH oxidase in neuronal growth cones. Expression of NOX1-NOX4 family members at the RNA and protein levels has been shown for various brain regions including cortex, cerebellum, hippocampus, as well as for peripheral ganglia (Sorce and Krause 2009; Hernandes and Britto 2012). NOX2 subunits have been localized in cultured neurons, such as hippocampal (Tejada-Simon et al. 2005; Park and Jin 2008), cerebellar granule (Coyoy et al. 2008), sympathetic (Tammariello et al. 2000; Hilburger et al. 2005), and dorsal root ganglion neurons (Cao et al. 2009). However, these studies did not investigate the detailed subcellular distribution of NADPH oxidase proteins. Here, we have shown that a NOX2-type NADPH oxidase is localized to the plasma membrane of Aplysia neuronal growth cones, exhibiting a similar but less homogenous plasma membrane distribution compared to the cell adhesion protein apCAM. Whether another type of NADPH oxidase is also present in Aplysia growth cones, remains to be investigated. NADPH oxidase inhibition with VAS2870 or celastrol impaired actin organization and dynamics as well as neurite outgrowth, confirming our previous findings (this study; Munnamalai and Suter 2009), while NADPH oxidase activation increased H2O2 levels in the growth cone P domain (Figs. 1, 2; Figure S1). The cytosolic subunit p40phox exhibited relatively little colocalization with NOX2 in unstimulated growth cones, but had significant colocalization with F-actin bundles (Fig. 3). We observed variations in p40phox levels between growth cones, which correlated with F-actin content (Fig. 4). Cytochalasin B significantly reduced the amount of p40phox in the growth cone periphery (Fig. 5), while cell fractionation and actin coimmunoprecipitation experiments confirmed a partial association of p40phox with F-actin (Fig. 6). Cell fractionation studies were performed on total CNS lysates suggesting that the differential association of NOX2 and p40phox not only occurs in the growth cone but also in other regions of the cell. Differential localization of NOX2 and p40phox in other compartments of unstimulated neurons is further supported by confocal imaging of the neuronal cell body (Figure S4). Finally, we have found increased p40phox/NOX2 colocalization at adhesion sites during apCAM-evoked neuronal growth (Fig. 7). In summary, these findings provide evidence for a bidirectional functional relationship between NADPH oxidase activity and the actin cytoskeleton in neuronal growth cones, which may contribute to the control of neurite outgrowth.
Very little is known about NADPH oxidase signaling in neuronal development, particularly in axonal growth and guidance. Previous work from our laboratory and others suggested a role for NADPH oxidase-derived ROS in regulating neurite outgrowth (Suzukawa et al. 2000; Ibi et al. 2006; Munnamalai and Suter 2009). Tail wound-derived H2O2 can promote regeneration of sensory axons in Zebrafish embryos (Rieger and Sagasti 2011). Furthermore, we have shown that NADPH oxidase inhibition in Aplysia growth cones results in reduced F-actin content (this study; Munnamalai and Suter 2009). ROS have also been implicated in Rac1-dependent Ca2+ release in Aplysia growth cones (Zhang and Forscher 2009). Because of the short half-life and high reactivity of ROS, target proteins need to be close to the ROS source to establish signaling specificity (Ushio-Fukai 2006). Consistent with this idea, we found NADPH oxidase localized in the growth cone periphery at sites where actin assembly occurs. This study is also interesting from an evolutionary point of view. While ancestral NOX2, p22phox , p47phox , and p67phox homologs have been identified for several invertebrate species including Monosiga brevicollis, Nematostella vectensis, Strongylocentrotus purpuratus, as well as the snail Lottia gigantea, no p40phox homolog has been identified in these species thus far, which has led to the conclusion that p40phox may have evolved later with the chordata (Kawahara et al. 2007; Sumimoto 2008). Very recently several predicted Aplysia NADPH oxidase-like sequences have been released to the NCBI database, including NOX2-like (XP_005090645), NOX5-like XP_005101964), dual oxidase 1 (DUOX1)-like (XP_005108327), and DUOX2-like (XP_005106629). Our results suggest that a NOX2-type NADPH complex with all membrane and cytosolic subunits including the regulatory subunit p40phox already evolved with the appearance of the molluscan clade. Other available antibodies against NOX2-type subunits did not detect proteins with high enough specificity (data not shown); thus, because of this limitation we were unable to detect other subunits of the NOX2-type complex in Aplysia growth cones thus far.
Our quantification of H2O2 levels following PKC activation revealed that the NADPH oxidase is functionally active in growth cones. VAS2870 abolished the PDBu-stimulated increase in H2O2 levels in the growth cone periphery within 30 s (Fig. 2i). Treatment of growth cones with VAS2870 alone for 15 min resulted only in a modest decrease of H2O2 levels (Fig. 2h). These findings are in line with another study demonstrating that VAS2870 blocked the increase of ROS levels in endothelial cells activated by oxidized low-density lipoprotein, but had minimal effect on the basal ROS levels without stimulation (Stielow et al. 2006). There are several possible reasons why VAS2870 treatment alone did not lower basal H2O2 levels more. First, NADPH oxidases are not the only source for superoxide/hydrogen peroxide. Second, we are measuring hydrogen peroxide in this assay and not superoxide, which is the primary product of NADPH oxidase. Third, ROS are highly reactive, act locally, and are tightly regulated. Finally, the dye used to detect hydrogen peroxide may not be able to detect small ROS changes. Thus, it might be difficult to optically detect local changes in ROS levels induced by NADPH oxidase inhibition with sufficient temporal resolution, despite the fact that such changes can affect cellular physiology and morphology.
Cytoskeletal association of regulatory NOX subunits has been reported for p40phox, p47phox, and p67phox in non-neuronal cells (Nauseef et al. 1991; El Benna et al. 1994, 1999; Li and Shah 2002). Furthermore, p40phox and p47phox associate with the neutrophil cytoskeleton through a PX-domain-dependent interaction with the actin-associated protein moesin (Wientjes et al. 2001; Zhan et al. 2004), while p40phox also interacts with the actin-binding protein coronin (Grogan et al. 1997), as well as directly with the actin cytoskeleton (Shao et al. 2010). These findings suggested that actin association of cytosolic NADPH oxidase subunits may have a regulatory function by keeping the complex in an inactive state until a stimulus releases the subunits from the cytoskeleton during the activation process. In agreement with this idea, COS-7 expression experiments revealed that p40phox is associated with actin and moesin, and that a mutated form of p40phox shows higher actin association and inhibits NOX2 activity (Chen et al. 2007). Furthermore, the actin-binding protein cortactin has been suggested to regulate NOX2 activation in lung endothelial cells trough an interaction with p47phox (Usatyuk et al. 2007). Thus, actin rearrangements could also deliver cytosolic NADPH oxidase subunits to the plasma membrane. Does NOX2 localization show actin dependence? In migrating endothelial cells, NOX2 colocalizes with F-actin along the leading edge in an IQ motif containing GTPase activating protein 1 (IQGAP1)-dependent manner, while manipulating the actin cytoskeleton with latrunculin and jasplakinolide alters NOX2 localization (Ikeda et al. 2005). Our growth cone studies also suggest a certain actin dependence of NOX2 localization as indicated by the reduction of NOX2 following cytochalasin treatment and the colocalization of NOX2 with actin along the leading edge and in transition zone ruffle/intrapodia (Fig. 5g and h). Thus, a fraction of NOX2 along with p40phox appears to undergo retrograde flow together with F-actin. In summary, our present work in neuronal growth cones is consistent with previous findings made with leukocytes and endothelial cells indicating that the actin cytoskeleton may play a role in NADPH oxidase activation through localization of its subunits.
On the other hand, NADPH oxidase-derived ROS regulate actin organization and dynamics, growth cone motility, and neurite growth (this study; Munnamalai and Suter 2009) and actin organization and migration of non-neuronal cells (Nimnual et al. 2003; Wojciak-Stothard et al. 2005; Schroder et al. 2007; Kim et al. 2009, 2011; Kuiper et al. 2011). Thus, together with these findings, our current study suggests an interesting bidirectional relationship between NADPH oxidase and F-actin. We propose that the cytosolic NADPH oxidase subunits such as p40 are associated with actin structures in unstimulated growth cones. Upon growth cone stimulation by guidance cues, cytosolic subunits p47phox, p67phox, p40phox, and Rac1 translocate to the plasma membrane (either with or without F-actin) and activate the membrane-bound subunits NOX2/p22phox. NADPH oxidase-produced superoxide (or most likely hydrogen peroxide) diffuses back into the growth cone for local regulation of the cytoskeleton and related directional growth (see graphical abstract of the online version). In agreement with this hypothesis, we found increased p40phox/NOX2 levels and colocalization at growth cone contact sites with apCAM beads (Fig. 7) and interacting growth cones (arrow in Fig. 3e), where apCAM exhibits increased density as well (Thompson et al. 1996; Suter et al. 1998). These results indicate that apCAM clustering can trigger actin remodeling as well as NADPH oxidase activation related to neurite outgrowth. The details of the functional relationship and order of events of apCAM clustering, actin reorganization, and NADPH oxidase activation remain unclear at this point. apCAM clustering could first activate NADPH oxidase followed by ROS-mediated actin reorganization; or apCAM-mediated actin remodeling could activate NOX2 by translocating cytosolic subunits close to plasma membrane. Thus, additional studies are needed to address the important question whether NADPH oxidase activation plays a functional role in axonal growth triggered by apCAM or other guidance cues.
Acknowledgements and conflict of interests disclosure
We thank members of the Suter laboratory for their comments on this manuscript. PF-6 AM was kindly provided to us by Dr. Christopher Chang and Vivian Lin, University of California, Berkeley CA, USA. We also thank Kelsey Martin, Samuel Schacher, and Eric Kandel for providing 4E8 hybridoma cells to produce 4E8 antibody. This work was supported by grants from the NIH (R01 NS049233 to D.M.S. and P20 GM103500 to M. Q.), NSF (1146944-IOS to D.M.S.), and the Bindley Bioscience Center at Purdue University (D.M.S.). V. M. was partially supported by fellowships through the PULSe and PUN Integrative Neuroscience graduate programs at Purdue University. The authors have no conflict of interest to declare.
All experiments were conducted in compliance with the ARRIVE guidelines. The authors have no conflict of interest to declare.