Alsin and SOD1G93A Proteins Regulate Endosomal Reactive Oxygen Species Production by Glial Cells and Proinflammatory Pathways Responsible for Neurotoxicity*

Recent studies have implicated enhanced Nox2-mediated reactive oxygen species (ROS) by microglia in the pathogenesis of motor neuron death observed in familial amyotrophic lateral sclerosis (ALS). In this context, ALS mutant forms of SOD1 enhance Rac1 activation, leading to increased Nox2-dependent microglial ROS production and neuron cell death in mice. It remains unclear if other genetic mutations that cause ALS also function through similar Nox-dependent pathways to enhance ROS-mediate motor neuron death. In the present study, we sought to understand whether alsin, which is mutated in an inherited juvenile form of ALS, functionally converges on Rac1-dependent pathways acted upon by SOD1G93A to regulate Nox-dependent ROS production. Our studies demonstrate that glial cell expression of SOD1G93A or wild type alsin induces ROS production, Rac1 activation, secretion of TNFα, and activation of NFκB, leading to decreased motor neuron survival in co-culture. Interestingly, coexpression of alsin, or shRNA against Nox2, with SOD1G93A in glial cells attenuated these proinflammatory indicators and protected motor neurons in co-culture, although shRNAs against Nox1 and Nox4 had little effect. SOD1G93A expression dramatically enhanced TNFα-mediated endosomal ROS in glial cells in a Rac1-dependent manner and alsin overexpression inhibited SOD1G93A-induced endosomal ROS and Rac1 activation. SOD1G93A expression enhanced recruitment of alsin to the endomembrane compartment in glial cells, suggesting that these two proteins act to modulate Nox2-dependent endosomal ROS and proinflammatory signals that modulate NFκB. These studies suggest that glial proinflammatory signals regulated by endosomal ROS are influenced by two gene products known to cause ALS.

Amyotrophic lateral sclerosis (ALS) is a lethal degenerative neurological disorder characterized by progressive degeneration of motor neurons in the brain and spinal cord (1,2). The majority of ALS patients have onset of disease between 40 and 50 years of age and about 50% of patients die within 3 years. The majority of ALS cases are categorized as sporadic with no family history of disease. In this context, the causative genes and environmental factors that initiate the disease process remain poorly defined. Only ϳ10% of ALS cases have a clearly inherited genetic component and hence are classified as familial ALS (1,2).
The best-characterized forms of familial ALS include those caused by mutations in the gene encoding Cu/Zn-superoxide dismutase (SOD1) 2 (3). Approximately 20% of familial ALS cases are caused by a variety of dominant SOD1 mutations (1,3). There remains great uncertainty as to the primary mechanism(s) by which mutant SOD1 leads to pathology observed in ALS (1,4). Proposed mechanisms include toxicity associated with misfolding of mutant SOD1, such as ER stress and inhibition of the proteasome, enhanced proinflammatory ROS production, altered axonal transport, excitotoxicity caused by glutamate mishandling, and mitochondrial damage (1,4). Relevant to the studies in this report are findings demonstrating that SOD1 mutations induce NADPH oxidase-dependent ROS production in microglia of SOD1 G93A mice leading to motor neuron death (5,6). These studies have demonstrated that deletion of the Nox2 isoform of NADPH oxidase, and to a lesser extent also the Nox1 isoform, can prolong survival in SOD1 G93A transgenic mice. The importance of SOD1 in regulating cellular ROS production was first revealed by studies demonstrating that SOD1 can directly associate with endosomal Rac1 to regulate its activity (7). Rac1 is an essential activator of several NADPH oxidases and SOD1 binding to Rac1 slows the hydrolysis of GTP bound to Rac1 in a redox-dependent manner. Thus, SOD1 association with Rac1 is proposed to be a redox-dependent sensor for regulating redox-active NADPH oxidase containing endosomes (called redoxosomes) (7,8). Redoxosomes are important signaling endosomes that regulate proinflammatory receptor signals in a redox-dependent manner through NADPH oxidases (8 -11). Redoxosomes are a subpopulation of early endosomes that produce ROS and have been shown to contain early endosomal markers (Rab5 and EEA1), redox effectors (Nox2 or Nox1, Rac1, p47 phox , p67 phox , and SOD1), and certainly ligand-activated cytokine receptors (TNFR1 or IL-1R1) (8 -11). ROS produced by the redoxosome facilitate the redox-dependent recruitment of TRAFs to TNFR1 and IL-1R1 and in this manner facilitate proinflammatory signaling such as NFB activation (8 -11). Interestingly, mutant forms of SOD1 have enhanced redox-independent binding to Rac1 and this has been proposed as a mechanism for enhanced ROS production in microglia of ALS mice (7,12).
Seven known NADPH oxidase catalytic subunits exist (Nox1, Nox2 gp91phox , Nox3, Nox4, Nox5, Duox1, and Duox2) (13,14). The most widely characterized of these is phagocytic gp91 phox (also known as Nox2). Nox2 is also expressed in microglia (6) and a variety of other nonphagocytic cell types. Rac, a small GTPase, is an essential activator of Nox1 and -2, and along with several other subunits (p22 phox , p40 phox , p47 phox , p67 phox , NoxO1, and NoxA1) can act to promote Nox complex activation in a cell type-specific fashion (13,14). In certain cell systems, Rac1 has also been shown to be required for ROS production by Nox4 (15)(16)(17). Because  It is presently unclear if familial and sporadic forms of ALS have common or overlapping molecular mechanisms of disease pathogenesis. A recent genomewide association study in sporadic ALS patients has begun to shed light on this topic (18). Several genes that regulate endosomal trafficking, Rac1, and NADPH oxidases were identified in this study, including Nox4, TIAM2, IQGAP2, PTPRT, RAP1GAP, and MAGI2 (12). Interestingly, the ALS2 gene product alsin, which when mutated causes juvenile ALS, has also been shown to influence endosomal trafficking and Rac1 activity (19 -23). Alsin appears to serve as a GEF for Rab5 and an effector of Rac1 GTPase activity (24 -26). These findings are of considerable interest because SOD1 also regulates Rac1 GTPase and NADPH oxidase activity in Rab5-bound early endosomes (7,8,11,12). Both the Rab5-GEF and Rac1-effector functions of alsin appear to influence endocytic mechanisms and endosomal dynamics (20,26) and alsin appears to protect from motor neuron degeneration in certain SOD1 mutant mice (20) and motor neurons expressing SOD1 mutants in culture (27,28). Given the association of Nox1, Nox2, and Nox4 with disease progression in ALS mice (5,6) and humans (18), these findings suggest the intriguing hypothesis that alsin and SOD1 both influence the dynamics of Rac1-dependent, NADPH oxidase-mediated, ROS production by redoxosomes that may impact proinflammatory signaling in ALS. In support of this hypothesis, alsin has been shown to bind three components of the redoxosome (Rac1, Rab5, and SOD1).
To test this hypothesis, we investigated whether alsin expression influences SOD1 G93A -mediated ROS production by glial cells. Three NADPH oxidases were evaluated as sources of cellular ROS (Nox1, -2, and -4) using shRNA knockdown, based on their association with disease severity in ALS models. Findings from our studies demonstrated that wild type alsin attenuates SOD1 G93A -mediated Rac1 activation, ROS production by Nox2, NFB activation, and TNF␣ secretion by glial cells and protects neurons from toxicity in co-culture studies. SOD1 G93A expression enhanced TNF␣-dependent redoxosomal ROS production by Nox2 and this was attenuated by alsin expression. Taken together, our results suggest a potential role for alsin in regulating redox-dependent proinflammatory signals via redoxosomes that are enhanced by SOD1 G93A .
NFB and NADPH Oxidase Activity Assays-NFB transcriptional activity was assessed using the previously described NFB-inducible luciferase reporter vector (Ad.NFBLuc) (30). Luciferase activity was assessed with 2 g of cell lysate. NADPH-dependent superoxide production (i.e. NADPH oxidase activity) was analyzed by a lucigenin-based chemiluminescent assay. Prior to the initiation of the assay, 1 g of endomembrane protein (100,000 pellet, see below) was combined with 5 M lucigenin (Sigma) in PBS and incubated in the dark at room temperature for 10 min. The reaction was initiated by the addition of 100 M NADPH (Sigma), and changes in luminescence were measured over the course of 3 min. The slope of the luminescence curve (relative light units per minute) was used to calculate the rate of ROS formation as an index of NADPH oxidase activity (9,11). In the absence of NADPH, background levels of lucigenin-dependent luminescence were always Ͼ1,000-fold less than maximally induced values in the presence of NADPH. Additionally, background levels of luminescence in the absence of NADPH did not significantly vary between samples.
Cellular Endomembrane Collection and Western Blotting-Cells were scraped into ice-cold PBS and centrifuged. Cell pellets were resuspended in 0.5 ml of homogenization buffer (0.25 M sucrose, 10 mM triethanolamine, 1 mM phenylmethylsulfonyl fluoride, and 100 g/ml of aprotinin). Samples were lysed by nitrogen cavitation (650 p.s.i. for 5 min) and then centrifuged at 3,000 ϫ g at 4°C for 20 min. The supernatant was designated post-nuclear supernatant. The postnuclear supernatant was subsequently centrifuged at 100,000 ϫ g for 2 h at 4°C to pellet the endomembranes (100,000 pellet), which were then resuspended in 100 l of homogenization buffer and used immediately for ROS production assays or frozen for later analysis. Western blotting was performed using standard protocols and protein concentrations were determined using the Bio-Rad protein quantification kit. Immunoreactive proteins were detected using an Odyssey Infrared Imaging System (LI-COR Biotech). Anti-alsin antibody was purchased from Sigma.
Redoxosome Detection-Intra-endosomal ROS production was detected using OxyBURST Green dihydro-2Ј,4,5,6,7,7Јhexafluorofluorescein (H 2 HFF)-BSA (Invitrogen). Stock solutions (1 mg/ml) were generated immediately prior to use by dissolving H 2 HFF-BSA in PBS under nitrogen and protected from light. Cells were preincubated in the presence of 50 g/ml of OxyBurst Green H 2 HFF-BSA and TNF␣ (0.1 ng/ml) for 20 min at 37°C and then fixed in 4% paraformaldehyde. Cells were then immediately imaged using a Leica spinning-disc fluorescence microscope. For samples infected with adenoviral vectors, the infections were done 36 h prior to assays. When SOD/ catalase proteins were added to the medium, this was done at a protein concentration of 0.1 mg/ml at the time of TNF␣ treatment.
Detection of Cellular ROS Using DHE-Eight hours prior to addition of DHE (dihydroethidium), cell cultures were switched to a serum-deprived (0.2% serum) medium. Cell samples were stained with DHE (1 M) for 30 min in the dark at 37°C (34). Cells were then rinsed with PBS and evaluated using a spinning-disc fluorescence microscope. For flow cytometry, cells were harvested by trypsinization and filtered through a 70-m nylon cell strainer (Falcon) and then assessed by FAC-Scan flow cytometer (BD Biosciences).
Collection of Conditioned Glial Cell Medium-Glial cells cultured in DMEM, 10% FBS were infected with Ad.G93A, Ad.Empty, or Ad.shNox vectors as indicated. At 8 h postinfection, virus-containing media were discarded, glial cells were washed with fresh media twice and subsequently cultured in fresh media. At 48 h postinfection, culture media were collected as conditioned glial cell medium and immediately applied to naive neuronal cell or glial cell cultures.
Cell Labeling with DiO or DiI-Glial or neuronal cells were incubated with Vybrant DiO or Vybrant DiI (Invitrogen) for 30 min at 37°C at a final concentration of 1 M (35). To gain a universal cell labeling, this labeling step was performed every other day for a total of 3 times before adenoviral infections and combining the two separately labeled cell types into one co-culture. Co-cultures were initiated by mixing 1.0 ϫ 10 6 DiI-labeled glial cells and 0.5 ϫ 10 6 DiO-labeled neurons. At 12 h and 5 days after initiating the co-cultures, cells were harvested by trypsinization, filtered through a 70-m nylon cell strainer (Falcon), and the cell numbers were assessed with a BD Biosciences LSR II Violet Flow Cytometer with 405-, 488-, 561-, and 639-nm lasers.
Cytokine Measurements and Rac1 Activity Assays-The concentrations of cytokines IL-6 and TNF␣ in the culture medium were determined with the mouse IL-6 ELISA kit or the mouse TNF␣ ELISA kit (BD Biosciences), respectively. Rac1 activity assays were performed using a Rac1 G-LISA TM Activation Assay Kit (Cytoskeleton) following the manufacturer's protocol.

Alsin Modulates ROS Production by SOD1 G93A Expressing
Glial Cells-We and others have previously demonstrated that expression of SOD1 G93A enhances ROS production in spinal cords of ALS mice (5)(6)(7). The source of this ROS production by SOD1 G93A or SOD1 G37R expressing glial cells has been predominantly thought to be Nox2 (5)(6)(7)36). However, genomewide association studies in sporadic ALS patients have also linked Nox4 to the progression of disease (18). To this end, we were interested in systematically evaluating whether Nox1, Nox2, and/or Nox4 contribute to excessive ROS production by SOD1 G93A expressing glial cells and whether alsin (another gene known to cause juvenile ALS) also regulates ROS through any of these Nox isoforms. To approach this question, we used recombinant adenoviruses capable of overexpressing SOD1 G93A and alsin, or inhibiting Nox1, -2, and -4 through expressed shRNAs. As anticipated, expression of SOD1 G93A in a glial cell line (MO59J) led to a significant increase in ROS production as detected by DHE fluorescence and FACS analysis (Fig. 1A). SOD1 G93A -induced ROS production was inhibited by the addition of two Nox inhibitors, diphenylene iodonium and apocynin (Fig. 1B). Analysis of NADPH-dependent ROS production by the endomembrane fraction, using the chemiluminescence lucigenin probe, substantiated these findings demonstrating that SOD1 G93A expression induced ROS (likely superoxide based on the specificity of lucigenin) production that was inhibited by treatment of cells with diphenylene iodonium and apocynin (Fig. 1C).
We next sought to evaluate whether Nox1, -2, and/or -4 was responsible for the enhanced ROS production in SOD1 G93A -expressing glial cells using expressed shRNAs. Analysis of Nox1, -2, and -4 mRNA levels in MO59J cells using quantitative PCR demonstrated that only Nox2 and Nox4 were expressed at detectable levels. Expression of Nox2 or Nox4 shRNAs demonstrated significant and specific knockdown of their respective cellular target mRNAs by ϳ50% (supplemental Fig. S1). Because endogenous Nox1 mRNA was undetectable, we chose to utilize Nox1-shRNA as a negative control in selected studies. The ability of these Nox-shRNAs to attenuate SOD1 G93A -induced ROS production by glial cells was then evaluated by co-infection with the respective adenoviral vectors. Results from these experiments demonstrated that only Nox2-shRNA significantly inhibited ROS production (Fig. 1D); no significant decrease in ROS was seen with either Nox1-or Nox4-shRNA, although there was a minor decrease noted in Nox4-shRNA expressing cells.
Deletions in the alsin gene (ALS2), a large 1651-amino acid protein, causes a juvenile form of ALS (23). It has been hypothesized that alsin could potentially regulate Nox activity like SOD1 (12), but no direct information on this exists to date. We therefore sought to investigate whether alsin expression influences Nox-dependent ROS production in glial cells. Indeed, expression of full-length wild type alsin in glial cells significantly induced ROS production as detected by DHE in whole cells and by lucigenin with endomembranes ( Fig. 2, A-D). Surprisingly, coexpression of alsin with SOD1 G93A significantly reduced ROS production by glial cells from that induced by SOD1 G93A expression alone (Fig. 2). These results demonstrate for the first time that alsin has the ability to modulate glial cell ROS production in a fashion that is altered by the presence of SOD1 G93A ; alsin increases ROS production in otherwise healthy glial cells, but decreases ROS production in the presence of SOD1 G93A . In the fluorescent images, the mitochondria are stained green and the nucleus is stained blue. C, mouse glial cells were infected with Ad.G93A or Ad.Empty for 36 h prior to cell harvesting. A 1-g portion of each 100,000 endomembrane pellet was used to evaluate NADPH-dependent production of O 2 . by lucigenin-based chemiluminescence. D, mouse glial cells were infected with the indicated adenoviral vectors for 28 h prior to changing to serum-deprived (0.2% serum) medium for 8 h. Cells were then incubated with DHE (1 M) for 30 min at 37°C before FACScan flow cytometry. Data in all graphs represent the mean Ϯ S.E. with n ϭ 3; marked comparisons (* or #) demonstrate significant differences as determined by one-way analysis of variance followed by Student's t test (p Ͻ 0.05). . by lucigenin-based chemiluminescence. C, mouse glial cells were infected with the indicated vectors for 28 h prior to changing to serum-deprived (0.2% serum) medium for 8 h, and then incubated with DHE (1 M) for 30 min at 37°C. Cells were then assessed by FACScan. D, mouse glial cells were infected and treated as in C and quantified as either the percentage of total cell population (left y axis) or relative DHE intensity (right y axis). Data represent the mean Ϯ S.E. with n ϭ 3; marked comparisons (* or #) demonstrate significant differences as determined by one-way analysis of variance followed by Student's t test of the comparisons (p Ͻ 0.05).
Alsin Coexpression in SOD1 G93A -expressing Glial Cells Protects Neurons from Cell Toxicity in Co-culture-Motor neuron cell death is diagnostic for ALS pathology. Previous studies have demonstrated that glia expressing SOD1 ALS mutants secrete Nox2-dependent factors that kill motor neurons (36,37). To investigate the relevance of SOD1 G93A -induced glial ROS production to the mechanism of neuronal cell death in ALS, we developed a co-culture method using MO59J glial cells and the NSC-34 motor neuron-like cell (a hybrid cell line produced by fusion of motor neuron-enriched, embryonic mouse spinal cord cells with mouse neuroblastoma (38 -40)). To differentiate glial cells from neuronal cells in co-culture, two fluorescent lipophilic dyes, DiO (green) and DiI (orange), were chosen to mark membranes of each cell type. Each dye is highly fluorescent and photostable when incorporated into cell membranes. Following selective labeling of MO59J cells (with Dil) and NSC-34 cells (with DiO) for a week in culture, nearly all cells were labeled. No evidence for toxicity was noted from each of these dyes after prolonged culture (2 weeks). Co-culture for 1 week after labeling led to no detectable inter-cell translocation of the dyes (supplemental Fig. S2).
We next sought to investigate the influence of SOD1 G93Aexpressing MO59J glial cells (labeled with DiI) on co-cultured NSC-34 motor neuron-like cells (labeled with DiO). Glial cells were infected with Ad.SOD1 G93A or Ad.Empty viruses 36 h prior to co-culture. Five days after the initiation of co-culture, cells were assessed by fluorescence microscopy (Fig. 3A, top panel) or by flow cytometry (Fig. 3, A, bottom panel, and B and C) to count the neuronal (green) and glial (red) cells. From 12 h to 5 days of co-culture, both glial and neuronal cells proliferated in control conditions (i.e. Ad.Empty-infected glial cells) leading to a 2.1-and 4.5-fold amplification of these cell types, respectively, over the 5-day period (Fig. 3, B and C). By contrast, coculture of Ad.SOD1 G93A -infected glial cells with neuronal cells led to a significant reduction in the number of neuronal cells in the co-culture after 5 days, without a significant change in the number of glial cells (Fig. 3, B and C); neuronal cells cultured with Ad.SOD1 G93A -infected glial cells for 5 days expanded 54% less than the neuronal cells cultured with control Ad.Emptyinfected glial cells. These results demonstrate that SOD1 G93Aexpressing glial cells impair growth of neurons and/or lead to neuronal cell death. Although these results do not specifically distinguish between impaired neuronal growth and neuronal death, they do clearly demonstrate neurotoxicity caused by SOD1 G93A -expressing glial cells. Such findings are similar to previous studies using human ES cells (hESC)-derived glial cells/astrocytes expressing SOD G37R and SOD1 G93A in co-culture with hESC-derived motor neurons (36,37).
Having established our glial/neuronal cell co-culture system, we examined whether expression of alsin and/or Nox shRNAs in SOD1 G93A -expressing glial cells could influence the level of neurotoxicity observed in co-culture. To this end, labeled MO59J glial cells were infected with Ad.SOD1 G93A together with either Ad.Alsin or Ad.shNox2 and these cells were cultured with labeled NSC-34 motor neuron-like cells. Flow cytometry demonstrated that coexpression of alsin with SOD1 G93A in glial cells significantly increased the total number of neuronal cells in co-culture as compared with glial cells expressing SOD1 G93A alone ( Fig. 3D and supplemental Fig.  S3A). Similarly, coexpression of shNox2 with SOD1 G93A in glial cells also led to a significant increase in neuronal cell number in co-culture ( Fig. 3D and supplemental Fig. S3A). Interestingly, expression of alsin and shNox2 together with SOD1 G93A in glial cells did not have a synergistic effect on neurotoxicity ( Fig. 3D and supplemental Fig. S3A), suggesting that alsin and Nox2 act within the same pathway to inhibit neurotoxicity induced by SOD1 G93A -expressing glial cells. As a control for shRNAs that did not significantly change ROS production in SOD1 G93A -expressing glial cells, we also tested whether coexpression of Nox1-shRNA or Nox4-shRNA together with SOD1 G93A in glial cells would influence neurotoxicity in co-culture ( Fig. 3E and  supplemental Fig. S3B); as expected, no significant changes in neuronal cell survival/growth rate were seen in these studies. In comparison to earlier studies, these findings demonstrate a correlation between the level of ROS production by SOD1 G93Aexpressing glial cells (with and without alsin or Nox2shRNA) and the extent of neurotoxicity in co-culture. Additionally, neurotoxicity induced by SOD1 G93A -expressing glial cells was inhibited by treatment of cell co-cultures with the Nox2 inhibitor apocynin (supplemental Fig. S4). These findings provide further evidence for the Nox2 dependence of neurotoxicity.
Enhanced Production of TNF␣, but Not IL-6, by SOD1 G93Aexpressing Glial Cells Is Influenced by Nox2 and Alsin-Enhanced production of proinflammatory factors in the CNS, such as TNF␣ and IL-6, is thought to be centrally important to inflammation in a number of neurodegenerative diseases including ALS (4). For example, increased TNF␣ in the serum and CSF of ALS patients and mouse models has been correlated with the severity of disease (41)(42)(43). Similarly, enhanced IL-6 production in the serum and CSF has also been noted in ALS patients (44 -46), but others have also noted no differences (47,48). We hypothesized that enhanced production of these proinflammatory cytokines by SOD1 G93A -expressing glial cells might be linked to ROS production and the mechanism of neuronal cell death in co-culture. To this end, we examined the effects of SOD1 G93A , alsin, and/or Nox-shRNA expression in glial and neuronal cells on the ability of these cell types to produce TNF␣ and IL-6 in the culture medium.
Results from these experiments demonstrated that secretion of TNF␣ was significantly increased in both glial (4.9-fold) and neuronal (3.8-fold) cells expressing SOD1 G93A (Fig. 4, A and C). Interestingly, alsin expression alone also induced a modest and significant level of TNF␣ production from glial cells (2.5-fold, Fig. 4A), however, the low level induction in neuronal cells (2.3fold) did not reach significance (Fig. 4C). No differences in IL-6 production were noted in either glial or neuronal cells under these conditions (Fig. 4, A and C). We further examined whether alsin or Nox-shRNA expression could influence SOD1 G93A -dependent TNF␣ secretion. TNF␣ production by SOD1 G93A glial cells was significantly reduced by the expression of alsin or shNox2, but not shNox1 or shNox4 (Fig. 4B). None of these conditions altered IL-6 production by SOD1 G93A -expressing glial cells. These observed reductions in TNF␣ secretion by SOD1 G93A glial cells, facilitated by alsin or shNox2 expression, correlated with protection of neuronal cells in co-culture (Fig. 3D). Furthermore, neither alsin nor any of the Nox-shRNAs altered TNF␣ or IL-6 production in SOD1 G93A -expressing neuronal cells (Fig. 4D). Taken together, these studies demonstrate that both Nox2 and alsin can influence TNF␣ production by SOD1 G93A -expressing glial cells. It therefore appears probable that TNF␣ plays an important role in neurotoxicity resulting from co-culture with SOD1 G93A -expressing glial cells.

SOD1 G93A -expressing Glial Cells Secrete Factors That Activate NFB in Motor Neuron-like Cells-To better establish that
secreted factors (such as TNF␣) from SOD1 G93A -expressing glial cells influence survival of neuronal cells, we performed experiments with glial cell-conditioned media and evaluated both neuronal cell survival and activation of the proinflammatory transcription factor NFB. Indeed exposure of NSC-34 neuronal cells to conditioned medium from SOD1 G93A -expressing MO59J glial cells led to a significant reduction in the number of neuronal cells in culture over a 3-day period, as compared with conditioned medium from control cells (Fig. 5A). These results confirmed that cell-cell contact between glial and neuronal cells is not necessary for neuronal cell toxicity associated with SOD1 G93A -expressing glial cells. Furthermore, we found that coexpression of alsin or Nox2-shRNA, but not Nox4-shRNA, in SOD1 G93A -expressing glial cells diminished neuronal cell toxicity of conditioned medium, leading to a sig-nificantly larger neuronal cell population after 3 days in culture (Fig. 5A). These results are consistent with our findings that alsin and shNox2 reduce the production of ROS (Figs. 1 and 2) and TNF␣ (Fig. 4) in glial cell cultures and suggest that these redox-mediated proinflammatory changes may be responsible for neuronal cell toxicity.
Although activation of NFB has been linked to both survival and death signaling, in microglia and astrocytes NFB activation is associated with enhanced ROS production and the proinflammatory state that leads to motor neuron cell death in ALS (4). Furthermore, NFB activation by TNF␣ in motor neurons has been demonstrated to potentiate glutamate-induced spinal cord motor neuron death (49). Nox-dependent ROS-mediated signaling is also important for the activation of NFB by TNF␣ and IL-1 (9, 10, 29), two factors thought to be important in the progression of ALS disease. Therefore, we next examined the status of NFB transcriptional activity in both neuronal and glial cells exposed to conditioned medium harvested from SOD1 G93A -expressing glial cells and how modulation of alsin or Nox2 expression influenced this process. To this end, neuronal and glial cells were preinfected with a recombinant adenovirus containing an NFB-luciferase reporter, followed by exposure to conditioned medium harvested from glial cells infected with adenoviruses expressing SOD1 G93A , Nox2-shRNA, Nox4-shRNA, and/or alsin. Results from these experiments demonstrated that conditioned medium from SOD1 G93A -expressing glial cells significantly induced NFB transcriptional activity in neuronal cells by 24 h postexposure over that

. Conditioned medium collected from SOD1 G93A -expressing glial cells is neurotoxic and causes activation of NFB in neuronal and glial cells.
A, mouse glial cells were infected with the indicated vectors. At 8 h post-infection, virus containing media were washed off and switched to fresh media. At 48 h postinfection, glial cell-conditioned media were collected and exposed to neuronal cell cultures. The change in the number of neuronal cells after 3 days of exposure was then quantified. B, mouse glial cells were infected with either Ad.Empty or Ad.SOD1 G93A virus and conditioned media were generated as in A. These glial cell-conditioned media were then placed onto neuronal cells preinfected with Ad.NFBLuc. At each indicated time point post-glial cell medium exposure, neuronal cells were harvested for luciferase assay to determine NFB activation. C, mouse glial cells were infected with the indicated vectors and conditioned media were generated as in A. These glial cell-conditioned media were then placed onto neuronal cells preinfected with Ad.NFBLuc. After 24 h of exposure to glial cell-conditioned media, neuronal cells were harvested for luciferase assay. D, glial cell medium collected as in A were exposed to naive glial cells preinfected with Ad.NFBLuc. After 24 h of exposure to glial cell-conditioned media, glial cell cultures were harvested for luciferase assay. Data in all graphs represents the mean Ϯ S.E. with n ϭ 3; marked comparisons (*, #, and ∧) demonstrate significant differences using one-way analysis of variance followed by Student's t test of the comparisons (p Ͻ 0.05).
seen with control glial cell-conditioned medium (Fig. 5B). Coexpression of alsin or Nox2-shRNA with SOD1 G93A in glial cells significantly reduced the ability of glial cell-conditioned medium to activate neuronal NFB (Fig. 5C). Interestingly, conditioned medium from Nox4-shRNA expressing SOD1 G93A -glial cells attenuated NFB activation following exposure to neuronal cells, but this did not reach statistical significance (Fig. 5C and data not  shown). These findings demonstrate that glial cell secreted factors influence NFB activation in neuronal cells in a paracrine fashion. More importantly, they also demonstrate that alsin and Nox2 expression influence several features of SOD1 G93A -expressing glial cells in a similar fashion including: 1) glial cell ROS production, 2) toxicity to neuronal cells, 3) TNF␣ secretion by glial cells, and 4) paracrine activation of NFB in neuronal cells.
Excessive inflammation in ALS has been thought to involve amplification of a proinflammatory state exacerbated by excessive Nox-dependent ROS production (4,12). This interplay likely involves dysregulated production of inflammatory cytokines and neurotrophic factors that act in both a paracrine and autocrine fashion between activated microglia, activated astrocytes, and motor neurons. Our findings in Fig. 5, A-C, demonstrate the importance of alsin and Nox2 in regulating paracrine factors from SOD1 G93A -expressing glial cells and their effects on neuronal cell health. We sought to evaluate whether factors secreted by SOD1 G93A -expressing glial cells might also act in an autocrine fashion to activate other glial cells. To this end, we harvested conditioned media from glial cells expressing SOD1 G93A , Nox2-shRNA, Nox4-shRNA, and/or alsin, and exposed this media to naive glial cells expressing the NFBluciferase reporter. Results from these experiments demonstrated a striking similarity to that observed in neuronal cell experiments (Fig. 5D). Conditioned medium from SOD1 G93Aexpressing glial cells induced NFB activation in naive glial cells and this was attenuated by alsin or Nox2-shRNA, but not Nox4-shRNA expression. These findings clearly demonstrate that factors secreted by SOD1 G93A -expressing glial cells induced NFB activation in an autocrine fashion and that this pathway is modulated by alsin and Nox2. SOD1 G93A Expression in Glial Cells Enhances Redoxosomal ROS Production Induced by TNF␣-Our studies demonstrate that SOD1 G93A expression enhances ROS and TNF␣ production by glial cells. TNF␣ is known to induce Nox2-dependent ROS production in redoxosomes that harbor the ligand-bound TNF␣ receptor (TNFR1) (8,10). In this context, ROS production by the redoxosome signals redox-dependent activation of TNFR1-TRAF2 complex formation required for endosomal IKK-dependent NFB activation (10). SOD1 G93A has also been demonstrated to enhance endosomal ROS production by inhibiting GTP hydrolysis by Rac1 (an activator of Nox2) (7,12). Therefore, we hypothesized that enhanced redoxosomal ROS production in SOD1 G93A -expressing glial cells might amplify responsiveness to proinflammatory cytokines such as TNF␣. To investigate this hypothesis, we used a previously described method for detecting superoxide accumulation within intracellular vesicular compartments (11,50). This approach involved loading endosomes with the membrane-impermeable fluorochrome H 2 HFF (Oxyburst green) conjugated to BSA followed by fluorescent microscopy to detect endosomal ROS production.
Our initial studies, which evaluated changes in H 2 HFFdetectable endosomal ROS in SOD1 G93A -expressing and control glial cells, revealed no significant difference in endosomal ROS production (data not shown). Reasoning that detection limits for this technique likely require synchronization of receptor activation in redoxosomes, we altered our approach to compare the responsiveness of redoxosomes to low levels of TNF␣ (0.1 ng/ml) stimulation. To this end, we infected glial cells with Ad.SOD1 G93A or Ad.Empty virus and then treated these cells with TNF␣ and H 2 HFF-BSA prior to evaluating endosomal ROS at 20 min post-stimulation. This alternative approach demonstrated that indeed SOD1 G93A expression in glial cells significantly elevated both the number (5-fold) of TNF␣-induced redoxosomes and their fluorescent intensity of ROS production (1.8-fold), as compared with control TNF␣-stimulated cells (Fig. 6, A, B, and E). These findings are consistent with previous studies that localized SOD1 on the surface of redoxosomes as a Rac1bound redox sensor for Nox2-mediated ROS production (7,8). For example, ALS mutants of SOD1 bind more tightly to Rac1-GTP than wild type SOD1, and inhibit the redox-dependent hydrolysis of GTP-bound Rac1 on the surface of redoxosomes leading to enhanced Nox2-dependent ROS production (7). Additionally, wild type SOD1 recruits to the surface of TNF␣-activated redoxosomes containing TNFR1 to facilitate Rac1/Nox2-dependent ROS production and the redox-dependent recruitment of TRAF2 to the receptor complex, which is required for NFB activation (10). Thus, these new findings (Fig. 6, A, B, and E) demonstrate for the first time that ALS mutant SOD1 G93A functionally enhances redoxosomal ROS production in the context of a proinflammatory signaling pathway important in ALS (i.e. TNF␣). Such findings help to explain why SOD1 G93A expression in glial cells enhances Nox2-dependent proinflammatory signals that lead to NFB activation (Fig. 5D). SOD1 recruits to the surface of endosomes following stimulation with several proinflammatory cytokines (8,10,11,50) where it binds to Rac1 and regulates Nox-dependent activation in the redoxosome. Rac1-GTP is essential for Nox2 activation and SOD1 binding to Rac1-GTP decreases GTP hydrolysis in a redox-dependent fashion (7). In this manner, SOD1 binding to Rac1-GTP enhances Nox2-dependent ROS production in redoxosomes under reducing conditions, whereas under oxidizing conditions SOD1 uncoupling from Rac1-GTP leads to enhanced GTP hydrolysis and inactivation of the Nox2 complex (7,8). In this manner, SOD1 acts as a sensor for Nox-dependent redoxosomal ROS production (7,8). The finding that FALS mutants of SOD1 demonstrate diminished redox-dependent uncoupling from Rac1-GTP has been one mechanism proposed for enhanced proinflammatory ROS production in FALS (7, 12), a finding supported by our present studies. The mechanism by which alsin attenuates Nox2-dependent ROS production in SOD1 G93A -expressing glial cells (Fig. 2) remains unclear. Based on the previous observations that alsin is also an effector that binds Rac1 (12), we hypothesized that alsin may directly modulate proinflammatory Rac1-dependent ROS production by the redoxosome. To test this hypothesis we examined ROS production by TNF␣-stimulated redoxosomes in glial cells expressing alsin and/or SOD1 G93A . Interestingly, coexpression of alsin with SOD1 G93A in glial cells significantly attenuated the amount of endosomal ROS production following TNF␣ stimulation without altering the number of formed redoxosomes (Fig. 6, C and F). These findings suggest that alsin can modulate TNF␣-induced redoxosomal ROS that is enhanced by SOD1 G93A . Furthermore, expression of a dominant-negative form of Rac1 (N17Rac1) had an even stronger effect, by decreasing both the number of redoxosomes and the intensity of redoxosomal ROS production following TNF␣ stimulation of SOD1 G93A glial cells (Fig. 6, C and F), consistent with the fact that Rac1 is an essential activator of Nox2. Importantly, when purified bacterial SOD and bovine catalase proteins (which degrade superoxide and H 2 O 2 , respectively) were added to the medium with TNF␣ and H 2 HFF-BSA to degrade intra-redoxo-somal ROS as previously described (10,11), the intensity of endosomal H 2 HFF-BSA fluorescence and number of detectable redoxosomes were significantly reduced (Fig. 6, D and G). These results demonstrated that indeed the fluorescent signal detected by H 2 HFF-BSA was superoxide and/or H 2 O 2 derived. These findings demonstrate for the first time that alsin can functionally regulate SOD1 G93A /Rac1-dependent ROS production by TNF␣ signaling redoxosomes. These findings also help to explain how alsin can decrease ROS production (Fig. 2) and NFB activation (Fig. 5D) in SOD1 G93A -expressing glial cells.
Enhanced Redoxosomal ROS Production by SOD1 G93A -Expressing Glial Cells Is Proinflammatory-Our studies thus far have demonstrated that enhanced ROS production in SOD1 G93A -expressing glial cells correlates with enhanced TNF␣ secretion and NFB activation. Furthermore, enhanced ROS production by SOD1 G93A -expressing glial cells appears to originate from enhanced redoxosomal activity. To definitively demonstrate that increased redoxosomal ROS production in the presence of SOD1 G93A is indeed responsible for activation of NFB and TNF␣ expression, we evaluated whether quenching of ROS within the endosomal compartment would attenuate activation of NFB and secretion of TNF␣ in SOD1 G93Aexpressing glial cells. To this end, SOD1 G93A glial cells were incubated with purified SOD/catalase proteins under conditions that significantly neutralized redoxosomal ROS (Fig. 6, D and G) and the level of TNF␣ secretion and NFB activation was quantified. Results from these studies demonstrated that SOD/catalase endosomal loading significantly lowered TNF␣ levels in the medium 2.4-fold (Fig. 7A). Similarly, the rise in NFB activation induced by SOD1 G93A expression was diminished by ϳ50% in the presence of SOD/catalase in the medium (Fig. 7B). These findings provide strong support that redoxosomal ROS production induced by SOD1 G93A expression enhances the proinflammatory phenotype of glial cells by inducing NFB and TNF␣. SOD1 G93A and Alsin Both Induce Rac1 Activation and SOD1 G93A Expression Enhances Recruitment of Alsin to Endomembranes-Rac1 activation is required for Nox2-dependent superoxide production (11,51). Studies thus far suggested that Rac1 plays an important role in glial ROS production induced by SOD1 G93A expression (Fig. 6, C and F). Alsin has the unique ability to induce glial cell ROS in the absence of SOD1 G93A , whereas inhibiting ROS in the presence of SOD1 G93A (Figs. 2 and 6, C and F). We hypothesized that alterations in Rac1 activity induced by alsin and/or SOD1 G93A expression might control Rac1-dependent ROS production by glial cells. To test this hypothesis, we evaluated the abundance of activated Rac1 (Rac1-GTP) in glial cells expressing SOD1 G93A and/or alsin. Results indicated that SOD1 G93A and alsin expression can independently activate Rac1 in glial cells 5.1-and 3.4-fold, respectively (Fig. 7C). Consistent with the ability of alsin to reduce Nox2-dependent ROS in SOD1 G93Aexpressing glial cells, coexpression of alsin and SOD1 G93A in glial cells significantly attenuated Rac1 activation compared with SOD1 G93A expression alone (Fig. 7C). Together with earlier findings, these results provide strong support that alsin and SOD1 G93A both converge on Rac1 to regulate endosomal Nox2-dependent ROS production.
Previous studies have demonstrated that SOD1 recruits to endomembranes following activation of proinflammatory receptors such as TNFR1 and IL-1R (10,11,50). In this context, SOD1 binds Rac1 to regulate Nox-dependent endosomal ROS. ALS mutant SOD1 proteins that have a higher affinity for Rac1-GTP enhance endosomal ROS production (7,12). We reasoned that because alsin overexpression attenuates SOD1 G93A -induced Rac1 activation (Fig. 7C), alsin might recruit to the endosomal compartment in a SOD1 G93A -dependent manner to regulate Rac1/Nox2-dependent ROS production. To test this hypothesis, we evaluated the recruitment of alsin to the endomembrane fraction in cells overexpressing alsin and/or SOD1 G93A . Findings from these studies demonstrated that alsin recruitment to endomembranes was significantly increased 2.6-fold under conditions of SOD1 G93A expression (Fig. 7D). These and our earlier findings (Fig. 6, C and F) demonstrate that the ability of alsin to regulate redoxosomal ROS induced by SOD1 G93A correlates with changes in its location to the endosomal compartment.

DISCUSSION
Enhanced ROS production has been proposed to be an important pathophysiologic component for several types of neuronal degenerative diseases including ALS (4). This is perhaps not surprising because most neuronal degenerative diseases have significant components of inflammation. The challenge in understanding the importance of enhanced ROS production in ALS has been to determine whether primary defects in ROS regulation incite pathophysiologic events responsible for disease progression. The first report to implicate Nox2 as a source of altered ROS production in a FALS (SOD1 G93A ) mouse model, described by Wu and colleagues (6), was later substantiated by others (5). Additionally, work using hESC differentiated into astrocytes demonstrated that expression of SOD1 G37R -induced Nox2 protein production and ROSmediated killing of hESC-derived motor neurons in co-culture (36). Importantly, inhibition of Nox2 with apocynin rescued motor neuron survival in the presence of SOD1 G37R -expressing astrocytes (36). These in vitro studies by Marchetto and colleagues (36) were key to demonstrating that Nox2 plays an important intrinsic role within FALS astrocytes to cause paracrine killing of motor neurons. Together with the finding that SOD1 ALS mutants can also directly regulate Rac1 and Nox2dependent ROS production, these findings point to dysregulation of NADPH oxidases as an important potential mechanism in the pathology observed in FALS.
Studies have also demonstrated that Nox2 is up-regulated in spinal cord microglia of sporadic ALS patients (6), suggesting that the mechanisms of enhanced ROS production through NADPH oxidases may not be limited to SOD1 mutations found in FALS. The potential involvement of NADPH oxidases in FALS disease progression is supported by genomewide association studies demonstrating that Nox4 and several Nox regulators are closely linked to the disease (12,18). Furthermore, studies on the juvenile form of ALS caused by mutations in the ALS2 gene encoding for alsin (22) have demonstrated that alsin regulates Rac1 activity (21,25,26) and thus also has the potential to regulate NADPH oxidases that require Rac1 (i.e. Nox1, -2, and -4). These findings suggest that multiple independent ALS disease-causing and disease-associated genes may converge on regulatory pathways that influence NADPH oxidasedependent ROS production. In the present study, we sought to determine whether both SOD1 G93A and alsin may both regulate the activity of NADPH oxidases in glial cells and the effect of this regulation on neuronal survival.
As previously reported for SOD1 G37R -expressing astrocytes (36), our studies demonstrate that SOD1 G93A -expressing glial cells hyperactivate Rac1-and Nox2-dependent ROS production leading to an enhanced proinflammatory state (TNF␣ and NFB) and toxicity of a motor neuron-like cell line in co-culture. Nox1 and Nox4 appeared not to play a major role in these processes. The ability of SOD1 G93A to activate Rac1 is consistent with previous studies demonstrating that SOD1 binds Rac1-GTP and reduces the rate of GTP hydrolysis and that mutant forms of SOD1 have a higher affinity for Rac1-GTP (7).
Interestingly, expression of full-length wild type alsin together with SOD1 G93A in glial cells led to protection of neuronal cells in co-culture by decreasing proinflammatory activation of glial cells (i.e. Rac1 activation, Nox2-dependent ROS production, TNF␣ production, and NFB activation). These findings suggest that alsin plays a protective role and is consistent with reports demonstrating alsin knockdown induces motor neuron cell death (20,28) and that expression of full-length alsin protects against neurotoxicity caused by SOD1 mutations (27). Interestingly the RhoGEF domain of alsin was required for this protective effect (27), a finding consistent with alsin mediating protection through the Rho-GTPase Rac1.
Our studies demonstrate for the first time that alsin can attenuate proinflammatory pathways regulated by Nox2 in SOD1 G93Aexpressing glial cells and that this effect in turn influences survival of motor neuron-like cells. Previous studies have focused on the function of alsin in motor neurons and have demonstrated similar protective effects by alsin overexpression on SOD1 G93A -associated toxicity (27). The mechanism of alsin-mediated protection of motor neurons remains unclear, but previous studies have suggested that protection may be mediated through alsin binding to mutant SOD1 (27) and/or through Rac1 modulation (28). Our studies, demonstrating for the first time protective functions of alsin in glial cells, add to this growing body of literature. Interestingly, however, our studies also demonstrate that overexpression of alsin alone in glial cells (in the absence of SOD1 G93A ) also drives similar amplification of proinflammatory pathways as seen following expression of SOD1 G93A alone (alsin overexpression induced Rac1, glial cell ROS, TNF␣ production, NFB activation, and toxicity to motor neuron-like cells in co-culture). Thus, our findings suggest that alsin is not simply a protective modulator of SOD1 G93A toxicity in glial cells, but rather that both alsin and SOD1 can act to directly regulate proinflammatory signals by glial cells.
Alsin is known to localize to the endosomal compartment where it can serve as a GEF for Rab5, an early endosomal effector GTPase (19,25,28). Alsin is also an effector of Rac1 that has been shown to control endocytic mechanisms at the cell membrane (26,52) and endolysosomal trafficking (20). Thus, alsin is a key regulator of endosomal dynamics. We hypothesized that alsin may regulate proinflammatory pathways through redox-active signaling endosomes (i.e. redoxosomes) known to facilitate redox-mediated activation of proinflammatory receptors such as TNFR and IL-1R (8,10,11). Our findings demonstrating that alsin recruitment to endomembranes is enhanced by SOD1 G93A expression supports its potential function on redoxosomes. The recruitment of SOD1 to redoxosomes, following TNF␣ or IL-1␤ stimulation, is important for controlling Rac1-dependent NADPH oxidase activation, redox-dependent TRAF recruitment to receptor complexes, and ultimately activation of NFB (8). Our novel findings that SOD1 G93A expression induces redoxosomal ROS production by TNF␣, and that this ROS is attenuated by alsin overexpression, provide a mechanistic link for anti-inflammatory effects of alsin in the presence of SOD1 G93A . Thus, our findings are consistent with alsin being a modulator of proinflammatory Nox2-dependent redoxosomal activation.
In summary, our studies demonstrate that SOD1 G93A expression in glial cells modulates proinflammatory signaling through redoxosomes in a Nox2-dependent fashion. This elevated proinflammatory state in turn leads to secreted factors that are toxic to neurons. Alsin appears to directly attenuate glial cell-dependent neurotoxicity by reducing Nox2-mediated signaling by redoxosomes. This protective effect appears to be mediated by the ability of alsin to decrease Rac1 activation in the presence of SOD1 G93A . Such findings provide insights into potential common regulatory pathways controlled by NADPH oxidases that influence proinflammatory signaling and neurotoxicity in two independent genetic forms of ALS.