Contribution of NADPH oxidase to the establishment of hippocampal neuronal polarity in culture

Reactive oxygen species (ROS) participate in pathological and physiological aspects of neuronal functions. A basal level of ROS produced by the NADPH oxidase (NOX) complex is necessary for neurotransmission, learning and memory (Kishida et al., 2006; Knapp and Klann, 2002; Massaad and Klann, 2011; Nayernia et al., 2014; Serrano and Klann, 2004). The NOX family consists of NOX1–NOX5, and DUOX1 and DUOX2 (Bedard and Krause, 2007), with NOX1, NOX2 and NOX4 being the main enzymes expressed in the central nervous system (CNS) (Sorce et al., 2012). NOX2 interacts with five regulatory proteins – p22^phox, p40^phox, p47^phox, p67^phox (also known as CYBA, NCF4, NCF1 and NCF2, respectively) and Rac1 (Bedard and Krause, 2007; Lambeth, 2004; Nayernia et al., 2014). NOX proteins have been detected in several regions of the adult mouse brain (Serrano et al., 2003). Mutations in gp91^phox (also known as CYBB), p47^phox, p67^phox and p22^phox are linked to chronic granulomatous disease (CGD), which is associated with cognitive impairment (Pao et al., 2004).

Neurons are highly polarized cells that have two functionally independent compartments, the somato-dendritic region and the axon, that emerge during the establishment of neuronal polarity (Caceres et al., 2012; Dotti et al., 1988; Szu-Yu Ho and Rasband, 2011).

The actin cytoskeleton is essential for neuronal polarization (Bradke and Dotti, 1999; Stiess and Bradke, 2011). Thus, Rac1 and Cdc42, members of the small GTPase Rho family, promote neuronal polarization and axonal growth (Gonzalez-Billault et al., 2012). Oxidation of actin decreases its ability to polymerize (Hung et al., 2011,, 2010; Sakai et al., 2012; Terman and Kashina, 2013). However, inhibition of NOX reduces both the F-actin content at the growth cone and the retrograde actin flow in neurons, suggesting a crosslink between NOX and actin dynamics (Munnamalai and Suter, 2009; Munnamalai et al., 2014).

In this work, we studied the contribution of the NOX complex to the development of neuronal polarity. Inhibition of the NOX complex affected polarity acquisition and reduced the axonal length of cultured neurons. NOX inhibition also affected actin organization, and decreased both the filopodial dynamics and the activity of Rac1 and Cdc42. These findings suggest that physiological levels of ROS, which are maintained by NOX, are needed to support neuronal polarization in vitro.

To evaluate the contribution of the NOX complex in the establishment of neuronal polarity, we used genetic and pharmacological strategies. First, embryonic hippocampal neurons were transiently co-transfected, after plating, with GFP and the P156Q mutant of the regulatory subunit p22^phox (here denoted DNp22^phox), which has a dominant-negative effect on ROS production, affecting the NOX1–NOX3 enzymes (Kawahara et al., 2005). Transfected neurons were fixed after 24 h in culture (Fig. 1). DNp22^phox expression delayed neuronal polarity acquisition (Fig. 1A,B). To evaluate the contribution of NOX to axonal growth, neurons transfected with DNp22^phox were cultured for 3 days in vitro (DIV) and were then fixed and immunostained for MAP2 and Tau (also known as MAPT) (somato-dendritic and axonal markers, respectively) (Fig. 1C). DNp22^phox expression decreased the length of axonal, but not minor neurites (Fig. 1D,E). In addition, MAP2 was detected in axons after DNp22^phox expression (Fig. 1C), suggesting that NOX inhibition disrupted neuronal polarization. To confirm that DNp22^phox indeed reduced ROS content, neurons were co-transfected with the genetically encoded biosensor Hyper, which detects intracellular H₂O₂ (Lukyanov and Belousov, 2014), an indicator of NOX activity. DNp22^phox expression (48 h) significantly reduced H₂O₂ content compared with control neurons (Fig. 1F,G). The Hyper-H₂O₂ map revealed that the highest H₂O₂ production was at the periphery of the soma as well as at the axonal tip (Fig. 1F, arrows), whereas DNp22^phox expression abolished this pattern. Expression of DNp22^phox was confirmed in N1E115 cells and cultured neurons (supplementary material Fig. S2). Taken together, these results suggest that NOX inhibition alters neuronal polarity acquisition and axonal growth.

As a second strategy to reduce NOX activity, neurons were treated with NOX inhibitors at 6 h after plating. Those chosen were gp91 ds-tat (5 µM), a peptide that inhibits p47^phox association with gp91^phox (Rey et al., 2001), VAS2870 (5 µM), a molecule that blocks the assembly of the NOX complex (Altenhofer et al., 2012) and apocynin (100 µM), which blocks p47^phox translocation to the plasma membrane (Ximenes et al., 2007). Neurons were fixed at 18 h of culture to evaluate the development of neuronal polarity (Fig. 2A). Under these treatments, most of the neurons remained at stage 1 (Fig. 2B–D), which supports the idea that NOX inhibition modifies neuronal polarity acquisition. We used DCFH-DA (a probe to measure oxidative species; LeBel and Bondy, 1990) to check ROS content after NOX inhibition (Fig. 2E). To rule out any non-specific effects of gp91 ds-tat, we used a scrambled gp91 (scr) peptide, which neither affected ROS content nor inhibited neuronal polarity (Fig. 2B,E).

Next, we sought to study the contribution of NOX to axonal growth. Neurons were treated with gp91 ds-tat, gp91 scr, VAS2870 and apocynin at 18 h of culture, when neurons are already at stage 2 and only display minor neurites. Neurons were fixed at 2 and 3 DIV to quantify the length of the axon and minor neurites (supplementary material Fig. S1; Fig. 2F–H). After 3 DIV, most neurons were fully polarized (stage 3) (92%±1), but this percentage decreased after NOX inhibition (gp91 ds-tat, 62%±13; VAS2870, 35%±15, P<0.01 and apocynin: 30%±3.5, mean±s.e.m., P<0.01). The remaining neurons did not develop an axon, resembling stage 2 of polarity. Control neurons exhibited somatic MAP2 and axonal Tau segregation at 3 DIV (87%±3) (Fig. 2F). In contrast, Tau and MAP2 distribution was reduced to 38%±5 of neurons treated with gp91 ds-tat (P<0.01), 3%±3 with VAS2870 (P<0.001) and 8%±1 with apocynin (P<0.001) (Fig. 2F), indicating loss of polarity. Moreover, NOX inhibition reduced the axon length but not the minor neurite length (Fig. 2G,H). These results are consistent with DNp22^phox-dependent NOX inhibition.

The NOX complex has been detected in mouse adult brain and in mature cultured hippocampal neurons (Serrano et al., 2003; Tejada-Simon et al., 2005), but not in embryonic brain or developing neurons. We detected gp91^phox, p22^phox, p47^phox and p67^phox subunits by immunoblotting in embryonic (E18.5) hippocampus and cerebral cortex (Fig. 3A) and also in stage 2 and 3 cultured neurons (18 h and 48 h, respectively) (Fig. 3B). Rac1, another component of the NOX complex, is expressed in hippocampal neurons at these stages (Santos Da Silva et al., 2004). gp91^phox, p22^phox and p47^phox were detected by immunofluorescence both in the soma and in minor neurites of stage 2 neurons (Fig. 3C). Interestingly, NOX subunits were also detected at the axon and axonal tip at stage 3 (Fig. 3D), which suggests that local production of ROS might be involved in axonal growth. Thus, NOX subunits are expressed in a timely manner to support neuronal polarity acquisition.

Based on our findings, we hypothesized that the NOX complex affects actin dynamics during neuronal polarization. First, we measured the neuronal lamellar area at stage 1. A well-structured lamella is important because minor neurites and the axon will emerge from this region (Caceres et al., 2012). Neurons were transfected with GFP or co-transfected with DNp22^phox and GFP immediately after plating, and fixed after a short time in culture to measure the area of the lamella. The F-actin and tubulin cytoskeleton were detected with phalloidin–Alexa-Fluor-546 and β3-tubulin immunolabeling, respectively (Fig. 4A). DNp22^phox expression significantly reduced the lamellar area compared with control neurons (Fig. 4B). Moreover, gp91 ds-tat, VAS2870 and apocynin also reduced phalloidin labeling in neurons (Fig. 2A, arrows). These results are consistent with the finding that actin at the growth cone of Aplysia bag cells is disorganized after NOX inhibition (Munnamalai and Suter, 2009).

Second, based on the possible influence of NOX on the integrity of the actin cytoskeleton, we evaluated filopodial dynamics at the tip of the axon as a parameter for actin polymerization. Neurons were transfected with either the genetically encoded probe Lifeact, which allows visualization of actin polymerization in real time (Riedl et al., 2008) or co-transfected with Lifeact and DNp22^phox (Fig. 4C). The number, length and lifetime of filopodia at the axonal tip were reduced after DNp22^phox expression (Fig. 4D–F). These results suggest that adequate amounts of ROS are needed to maintain the dynamics of the actin cytoskeleton.

Third, and considering that both lamellar and filopodial dynamics were altered after NOX inhibition, we sought to measure Rac1 and Cdc42 activities after DNp22^phox expression. To this end, 1 DIV neurons were transfected with the Raichu FRET biosensors for Rac1 or Cdc42 in control and DNp22^phox expression conditions (Nakamura et al., 2006). Representative FRET maps for Rac1 and Cdc42 are shown to indicate their local activity (Fig. 4G–J). Quantification of FRET efficiency was performed at the soma and for the whole axon, and its proximal and distal segments. The expression of DNp22^phox decreased Rac1 FRET globally (Fig. 4H), which is consistent with the decrease in axonal length and lamellar area (Fig. 1D; Fig. 4A). Cdc42 FRET efficiency, in turn, was decreased only within the distal axon (Fig. 4J), the same region where we observed a decrease in filopodial dynamics (Fig. 4C–F). Filopodial dynamics can also be regulated by Arp2/3 (Spillane et al., 2011), supporting the idea that Rac1 is also involved in this process, which is consistent with the decrease in Rac1 FRET efficiency shown in Fig. 4G. These results suggest that NOX inhibition modifies actin dynamics by decreasing the activity of Rho GTPase proteins. However, F-actin is also regulated by post-translational modifications of actin monomers that depend on redox balance (Hung et al., 2011; Terman and Kashina, 2013). Further experiments are thus required to explore the possibility that these modifications are required as well as regulation of the Rho GTPase protein activity.

gp91^phox- and p47^phox-knockout mice have normal brains, cortex and hippocampus (Kishida et al., 2006). However, axonal elongation, dendritic arborization and synaptic development have not been explored in these mice, even though the long-term potentiation (LTP) response is impaired and CGD patients present cognitive impairments (Kishida et al., 2006; Pao et al., 2004). Polarity acquisition studies in vivo could provide clues about the loss in neuronal functions observed in these models.

In conclusion, we propose that ROS production by the NOX complex contributes to the establishment of hippocampal neuronal polarity and axonal growth in vitro through the regulation of Rac1, Cdc42 and actin cytoskeleton dynamics. These findings support the idea that physiological levels of ROS are indeed necessary for normal neuronal development and function.

Pregnant Sprague-Dawley rats were killed, embryos (E18.5) were removed and neurons were cultured according to Kaech and Banker (2006). All animal experiments were performed according to approved guidelines.

N1E115 cells (ATCC, VA) were cultured in Dulbecco's modified Eagle's medium (DMEM) with 5% fetal bovine serum (FBS) to check DNp22^phox expression.

Neurons were transiently transfected with Lipofectamine 2000 (Life Technologies, CA) in Neurobasal medium. After 2 h, neurons were supplemented with B27, Glutamax, sodium pyruvate and antibiotics. Experiments were performed 18–72 h after cDNA transfection.

Antibodies against gp91^phox (mouse, ab109366, lot YH081212C; 1:1000 for immunoblotting and 1:100 for immunofluorescence), p67^phox (rabbit, ab80897, lot GR23630-9; 1:500 for immunoblotting) and p22^phox (rabbit, ab75941, lot GR83982-1; 1:1000 for immunoblotting and 1:100 for immunofluorescence) were purchased from Abcam (MA). The antibody against α-tubulin (1:10,000, mouse) was from Sigma (MO), that against p47^phox (rabbit, sc-14015, lot A2113; 1:500 for immunoblotting and 1:100 for immunofluorescence) was from Santa Cruz Biotechnology (TX), and those against MAP2 (rabbit, 1:500) and Tau (mouse, 1:500) were from Merck Millipore (Darmstadt, Germany). Anti-β3-tubulin antibody (mouse, 1:1000) was from Promega (WI). For solutions and general considerations for immunoblotting and immunofluorescence experiments, please see Henriquez et al. (Henriquez et al., 2012).

Neurons (4×10⁴ cells/well) were cultured on glass coverslips. Immediately after plating, neurons were transfected with Hyper (Evrogen, Moscow, Russia), an intracellular and ratiometric sensor to detect local H₂O₂ production (Lukyanov and Belousov, 2014). Transfected neurons were excited at 488 and 405 nm and emission was collected at 505–530 nm. Fluorescence emission from excitation at 488 nm was divided by fluorescence emission at 405 nm excitation (488:405) as a measure of the H₂O₂ content (Belousov et al., 2006).

Neurons were incubated with 1 µM DCFH-DA (Sigma) for 20 min at 37°C to evaluate intracellular ROS levels. DCFH-DA detects intracellular oxidative species by increasing fluorescence emission after oxidation (LeBel and Bondy, 1990). Neurons were fixed and permeabilized as described previously (Henriquez et al., 2012). β3-tubulin immunofluorescence and the transient expression of the far-red fluorescent protein mKate2 (Evrogen, Moscow, Russia) were used to normalize DCFH-DA emission, similar to as described previously (Munnamalai and Suter, 2009).

Neurons (24 h in culture) were fixed and immunostained against β3-tubulin. Phalloidin–Alexa-Fluor-546 was incubated for 1 h at room temperature during secondary antibody incubation for F-actin detection. Binary masks of F-actin- and β3-tubulin-positive areas were generated to measure the lamellar area. The lamellar area of stage 1 neurons was defined as the area of F-actin minus the area of β3-tubulin (Laishram et al., 2009).

Neurons were transfected with the Lifeact–GFP biosensor, and imaging was carried out 18 h after transfection. Time-lapse images were taken every 30 s for 10 min to visualize filopodial dynamics. Later, the number, length and lifetime of filopodia were measured using Fiji-ImageJ (NIH, Bethesda). Protrusions shorter than 2 µm and longer than 15 µm were not considered for the analysis. The lifetime was defined as the time during which a filopodium emerges and disappears.

Neurons (4×10⁴ cells/well) were transfected with the Raichu-Rac1 and Raichu-Cdc42 FRET biosensors (provided by Alfredo Cáceres, IMMF, Córdoba, Argentina) to measure Rho GTPase activity. Raichu probe expression and FRET efficiency measurements were performed as described previously (Nakamura et al., 2006). Briefly, transfected neurons were excited at 450 nm, and emissions were collected at 460–490 and 505–530 nm (donor and acceptor emission wavelengths, respectively). The ratio of the acceptor to donor emission was established as the FRET efficiency. The FRET map was achieved by dividing the acceptor to donor ratio image by the binary mask of the same image. Measurement of FRET efficiency was carried out by selecting a region of interest at the soma, the whole axon or the proximal and distal axon.

Results are the mean±s.e.m. of at least three independent cultures. The number of neurons per experiment (n) is indicated in the figure legends. ANOVA, Dunnett's post-test and Student's t-test tests were carried out with the GraphPad Prism 5 software.

We thank James Bamburg (Colorado State University, Colorado, USA) for the P156Q p22^phox construct and Michael Handford for English editing of the manuscript.