A peroxisome deficiency–induced reductive cytosol state up-regulates the brain-derived neurotrophic factor pathway

The peroxisome is a subcellular organelle that functions in essential metabolic pathways, including biosynthesis of plasmalogens, fatty acid β-oxidation of very-long-chain fatty acids, and degradation of hydrogen peroxide. Peroxisome biogenesis disorders (PBDs) manifest as severe dysfunction in multiple organs, including the central nervous system (CNS), but the pathogenic mechanisms in PBDs are largely unknown. Because CNS integrity is coordinately established and maintained by neural cell interactions, we here investigated whether cell-cell communication is impaired and responsible for the neurological defects associated with PBDs. Results from a noncontact co-culture system consisting of primary hippocampal neurons with glial cells revealed that a peroxisome-deficient astrocytic cell line secretes increased levels of brain-derived neurotrophic factor (BDNF), resulting in axonal branching of the neurons. Of note, the BDNF expression in astrocytes was not affected by defects in plasmalogen biosynthesis and peroxisomal fatty acid β-oxidation in the astrocytes. Instead, we found that cytosolic reductive states caused by a mislocalized catalase in the peroxisome-deficient cells induce the elevation in BDNF secretion. Our results suggest that peroxisome deficiency dysregulates neuronal axogenesis by causing a cytosolic reductive state in astrocytes. We conclude that astrocytic peroxisomes regulate BDNF expression and thereby support neuronal integrity and function.

The peroxisome represents a ubiquitous and essential subcellular organelle engaged in a variety of metabolic pathways, including biosynthesis of ether-phospholipids, ␤-oxidation of very-long-chain fatty acid (VLCFA), 2 ␣-oxidation of branched-chain fatty acid, and degradation of D-amino acids (1). Among these metabolic pathway, there are a number of oxidases that use O 2 to oxidize substrates and generate H 2 O 2 , such as acyl-CoA oxidases (AOx) and D-amino acid oxidases (2). Catalase is a peroxisomal tetrameric matrix enzyme that catalyzes degradation of H 2 O 2 . The physiological importance of peroxisomal metabolisms is demonstrated by severe pathological manifestations in peroxisome biogenesis disorders (PBDs). PBDs are caused by the defect of PEX genes encoding peroxisome biogenesis factors, peroxins (Pex). Zellweger spectrum disorders (ZSDs), accounting for about 80% of PBD patients, are classified into three groups according to their clinical severity: Zellweger syndrome (ZS), neonatal adrenoleukodystrophy, and infantile Refsum disease (IRD) (3). Patients with ZS, the most severe PBD, manifest severe impairment in the central nervous system (CNS), such as migration defect of cortical neurons, abnormal morphology of Purkinje cells, and dysplasia of inferior olivary nucleus (ION) (3)(4)(5)(6). Several Pexdeficient ZS model mice show the impairments of the CNS, as observed in ZS patients (7)(8)(9)(10). Moreover, in the Nes-Pex5 Ϫ/Ϫ mouse, the dysfunction of peroxisomes in all neural cells, including neurons, oligodendrocytes, and astrocytes, gives rise to abnormal development and aberrant morphology of CNS (11,12). However, brain cell type-specific (i.e. projection neuron-, astroglia-, or oligodendrocyte-specific) knockout of Pex5 does not show abnormal CNS development (13,14), suggesting that supportive effects among different brain cell types are responsible for normal development in the mutant mice (13). Very recently, we reported that the Pex14 ⌬C/⌬C mouse, a ZS model mouse, shows up-regulation of brain-derived neurotrophic factor (BDNF) in the neurons of ION and the elevation of a truncated form of its receptor, TrkB-T1, on Purkinje cells in the cerebellum (10). The malformation of Purkinje cells in the Pex14 ⌬C/⌬C mouse is caused by a combination of elevated BDNF and prominent expression of TrkB-T1 (10). Astrocytes, the most abundant cell type in the CNS, are engaged in divergent metabolic reactions and neuronal development (15), and neuron-astrocyte interaction plays a pivotal role in CNS integrity (16). Therefore, a cell co-culture system composed of two distinct types of brain cells, including neurons and astrocytes, might serve as a potential way to address the pathogenic mechanisms underlying abnormal development of neuronal cells.
In the present study, to uncover the pathological mechanism underlying PBDs, we focused on searching for soluble factors, if any, that influence neuronal development, by using a co-culture system of primary hippocampal neurons with glial cells (17,18). We found that peroxisome deficiency in astrocytes elevated the expression and secretion of BDNF, leading to the axonal branching of hippocampus neurons. We also show that the cytosolic reductive condition, but not the defects of peroxisomal ␤-oxidation and plasmalogen biosynthesis, is involved in the up-regulation of Bdnf mRNA. Therefore, these results suggest a new function of astrocytic peroxisomes in regulating BDNF expression for the neuronal integrity.

Establishment of peroxisome-deficient cultured astrocytes
To identify soluble factor(s) regulating neuronal development, we attempted to establish a peroxisome-deficient RCR-1 cell line, an astrocyte-like cultured cell line derived from rat embryonic cerebellum (19). To abrogate peroxisome biogenesis in RCR-1 cells, we designed two types of dominant-negative forms of Pex proteins, including C-terminally truncated Pex5p   (20) fused to a yellow fluorescent protein Venus (21), termed Pex5p-DN, and the N-terminal soluble region (residues 1-93) of Pex14p containing the Pex5p-binding domain (22) fused to Venus, named Pex14p-DN (Fig. S1A). Both truncated forms were anticipated to suppress the targeting of the endogenous Pex5p-cargo complex to Pex14p, resulting in perturbation of the peroxisomal matrix protein import (Fig. S1B). We established three RCR-1 cell lines, each stably expressing Pex5p-DN (RCR-1/Pex5p-DN), Pex14p-DN (RCR-1/Pex14p-DN), and Venus (RCR-1/Venus). Peroxisomal matrix protein import was assessed by immunofluorescent microscopy with antibody against peroxisome-targeting signal 1 (PTS1), which recognizes a dozen peroxisomal proteins, including AOx and trifunctional protein (23). Import of PTS1 proteins and catalase was severely impaired by the expression of the truncated mutant of Pex5p or Pex14p, but not Venus (Fig. 1,  A and B). Because several PTS1 proteins are unstable in the cytosol of peroxisome biogenesis-defective cells (24,25), fluorescent intensity of PTS1 proteins appeared to be decreased in RCR-1/Pex5p-DN and RCR-1/Pex14p-DN cells (Fig. 1A). Subcellular fractionation analysis revealed that almost all catalase was localized in the cytosol (Fig. 1, C and D). Intraperoxisomal processing of AOx and alkyldihydroxyacetonephosphate synthase (ADAPS) was attenuated, and AOx A-chain and the larger precursor of ADAPS were detected both in cytosol and organelle fractions from RCR-1/Pex5p-DN and RCR-1/ Pex14p-DN cells (Fig. 1C). To verify whether organelle-associated AOx and ADAPS precursor were transported into peroxisomal matrix, we performed protease protection analysis. Proteinase K treatment of post-nuclear supernatant (PNS) fraction leads to the marked reduction of AOx A-chain and ADAPS precursor in the organelle fraction (Fig. S1C, lanes 10 and 16), suggesting that only a part of AOx and ADAPS was translocated into the peroxisomal matrix ( Fig. S1C and Fig. 1A (e and f)). In the presence of Triton X-100, AOx and ADAPS were digested by the protease treatment (Fig. S1C, lanes 5, 6, 11, 12, 17, and 18). Consistent with these results, marked reduction of plasmalogens (PlsEtn; Fig. 1E) and accumulation of VLCFA-containing phosphatidylcholine (VLCPC; Fig. 1F) were evident. Therefore, peroxisomal biogenesis and metabolism were attenuated in the RCR-1 cells stably expressing Pex5p-DN or Pex14p-DN.

Co-culture of peroxisome-deficient astrocytes with primary hippocampal neurons
To examine the effect of astrocytic peroxisome deficiency on neural development, we established a cell co-culture system ( Fig. 2A). Using this co-culture system, we analyzed neuronal morphology in the early developmental stage at 2 days in vitro (DIV) (26). Neurons cultured with RCR-1/Venus as feeder cells extended a single primary axon (Fig. 2B, a and b). By contrast, neurons cultured with RCR-1/Pex5p-DN or RCR-1/ Pex14p-DN as feeder cells developed aberrantly elongated collateral branches (Fig. 2B, c, d, g, and h) with tertiary and quaternary branches (Fig. 2B, e, f, i, and j, arrowheads). We also analyzed the axonal morphology with anti-Tau-1 antibody, which recognizes the primary axons (Fig. S2A, arrows). From the primary axon with higher fluorescent intensity of Tau-1 staining, axonal collaterals more frequently emerged in the neurons co-cultured with peroxisome-deficient RCR-1 cells (Fig. S2A, arrowheads). We performed statistical analyses on the morphology of the neurons. The neurons isolated from the other cells were selected, and their longest processes were considered axons. In the neurons co-cultured with RCR-1/ Pex5p-DN or RCR-1/Pex14p-DN, no significant difference in the axon length was observed (Fig. 2C). By contrast, the percentages of neurons with branched axons were increased (Fig.  2D). Elongation of the collaterals was also promoted (Fig. 2E) and collaterals per 100 m of axon length were more frequent (Fig. 2F) in the neurons co-cultured with peroxisome-deficient RCR-1 cells as feeder cells than in those with control feeder cells. The formation of axonal collaterals was also evoked in neurons cultured in conditioned medium (CM) that had been prepared from peroxisome-deficient astrocytes (Fig. 2, G-J). Therefore, the secretion of factor(s) required for neuronal axon development is most likely dysregulated in astrocytes defective in peroxisome biogenesis.
To investigate whether cytoskeletal structure in the neurons was affected, actin filaments and microtubules were stained with fluorescent phalloidin and antibody to ␣-tubulin, respectively. Fluorescent phalloidin staining indicated actin cytoskeleton protruding from the axon of neurons, upon their coculture with peroxisome-deficient RCR-1/Pex5p-DN cells, whereas there was no obvious difference in microtubule structure between normal and RCR-1/Pex5p-DN cells (Fig. S2B). Thus, reorganization of actin cytoskeleton was properly promoted in the neurons co-cultured with the RCR-1 cells defective in peroxisomal biogenesis.

Elevation of BDNF in peroxisome-deficient astrocytes
Of the molecules involved in axon development, we focused on neurotrophic factors, including nerve growth factor (NGF), BDNF, glia-derived neurotrophic factor (GDNF), neurotrophin 3 (NT-3), and NT-4 (27). We recently demonstrated that a defect of peroxisome biogenesis in a neuroblastoma cell line, SH-SY5Y, leads to the up-regulation of BDNF, but not other factors including NGF, NT-3, and NT-4 (10). To assess mRNA levels of neurotrophins, RCR-1 cells were cultured in Neurobasal medium, which was used in a co-culture system, for 2 days, and then total RNA was extracted. Similarly to SH-SY5Y cells, Bdnf mRNA, not Ngf and Nt-3 mRNA, was elevated in RCR-1/Pex5p-DN and RCR-1/Pex14p-DN cell lines, as assessed by RT-PCR (Fig.  3A). The expression of Gdnf and Nt-4 was below detectable levels (data not shown). In real-time RT-PCR, the expression of Bdnf was 3-4-fold increased in both types of peroxisomedeficient RCR-1 cell lines compared with control RCR-1 cells (Fig. 3B). The level of secreted mature BDNF from the peroxisome-deficient astrocytes was also increased (Fig. 3C). To investigate whether the elevated BDNF induces axon branching, CM from cells expressing recombinant BDNF (rBDNF) was diluted and added to the culture medium of primary hippocampal neurons (Fig. 3D). The number of axon collaterals was increased by the addition of rBDNF in a D, RCR-1 cells were treated with 10 g/ml digitonin, and free catalase activity was determined in the isotonic medium. Free catalase activity was shown as a percentage of the total catalase activity detected in the presence of 1% Triton X-100. (n ϭ 3). E and F, total amounts of plasmalogens (E) and VLCPC (F) are represented by taking as 1 the amounts in RCR-1/Venus cells (n ϭ 3). **, p Ͻ 0.01; ***, p Ͻ 0.001, by Dunnett's test compared with RCR-1/Venus. Error bars, S.D.
BDNF positively regulates axonal outgrowth via binding to TrkB receptor on the neuronal surface (27). Knockdown of TrkB alone (Fig. 4A) affected neither axon elongation nor collateral formation under the culture conditions with CM prepared from RCR-1/Venus (Fig. 4, B-E). By contrast, in the pres-ence of CM from peroxisome-deficient RCR-1 cells, the promotion of collateral formation, but not axon elongation, was markedly lowered in the primary neurons with a significantly reduced level of TrkB (Fig. 4, B-E). The collateral generation in the CM from peroxisome-deficient cells was also decreased by inhibiting BDNF targeting to the TrkB with the recombinant extracellular domain of the p75 neurotrophic receptor  and concentrated by ammonium sulfate precipitation, and BDNF was detected by SDS-PAGE and immunoblot analysis with an anti-BDNF antibody. A molecular size marker in kDa is shown on the right. The BDNF band was quantified as described under "Experimental procedures" (bottom, n ϭ 3). D-H, CM from nontransfected RCR-1 (Ϫ) cells and RCR-1 cells stably expressing rBDNF was collected. Serial 10-fold dilutions (10 Ϫ2 to 10 Ϫ4 ) of rBDNF-containing CM were made up by adding CM from RCR-1 (Ϫ), and each was added to the culture of primary hippocampal neurons (E18.5) for 2 DIV as in Fig. 2 (A and B). D, secreted rBDNF and BDNF secreted to the CM from RCR-1 (Ϫ), RCR-1/Venus, and RCR-1/Pex14p-DN were analyzed as in C. *, nonspecific band. BDNF levels relative to that in the CM from RCR-1 (Ϫ) are indicated at the bottom. The band at the lane "10 Ϫ2 dilution" was saturated, indicating that its actual level is higher than the measured value of "40". Molecular mass markers were loaded in lane M, and their migrations were dotted at 10 and 15 kDa (dots). E-H, statistical analyses of axonal morphology were performed as in Fig

Peroxisome-deficient reductive state abrogates BDNF pathway
(p75ECD-His, Fig. S3A), another receptor for BDNF (27) (Fig.  S3, B-E). Taken together, our results suggested that the increase in secreted BDNF from peroxisome-deficient astrocytes gives rise to aberrant branching of neuronal axons mediated by TrkB. Because p75ECD-His also binds to other neurotrophins, such as NGF, NT-3, and NT-4, in CM (27), we cannot exclude the possibility that inhibition of axonal elongation ( Fig. S3B) might be owing to the effect of p75ECD-His on these factors.

Involvement of cytosolic catalase in regulation of BDNF expression in RCR-1 cells
To investigate how deficiency in peroxisome biogenesis enhances the Bdnf expression, several peroxisomal metabolisms, including synthesis of plasmalogens and fatty acid ␤-oxidation, were assessed. Restoration of plasmalogens in RCR-1/ Pex5p-DN and RCR-1/Pex14p-DN by treatment with hexadecylglycerol (28) did not alter the elevated Bdnf expression (Fig. S4, A and B), suggesting that plasmalogens were not involved in the regulation of Bdnf expression. Next, the deficiency of peroxisomal fatty acid ␤-oxidation in RCR-1 cells was induced by knockdown of AOx, the first-step enzyme in the ␤-oxidization, using short hairpin RNA for the AOx gene (shAOx). AOx proteins were markedly reduced (Fig. S4C), resulting in the accumulation of VLCPC (Fig. S4D) in RCR-1 cells stably expressing shAOx. However, there is no increment of the Bdnf mRNA level in AOx-depleted cells (Fig. S4E), suggesting that the up-regulation of Bdnf mRNA level is not induced by the impaired peroxisomal fatty acid ␤-oxidation. Interestingly, Bdnf expression was lowered by treatment with 20 mM 3-amino-1,2,4-triazole (3AT), a catalase inhibitor, of the peroxisome-deficient RCR-1 cells, RCR-1/Pex5p-DN and RCR-1/Pex14p-DN (Fig. 5A). At 60 mM 3AT, BDNF mRNA levels in both types of peroxisome-deficient cells were further lowered to that in untreated control cells (Fig. 5A). To investigate whether catalase in peroxisome-deficient cells is involved in the elevation of Bdnf mRNA, catalase was knocked down by siRNA treatment (Fig. 5E). In peroxisome-deficient cells, although catalase expression level was reduced, it remained in the cytosolic fraction (Fig. S5, lanes 7 and 11). Real-time PCR analysis revealed that the knockdown of catalase decreased the mRNA level of Bdnf in RCR-1/Pex5p-DN and RCR-1/ Pex14p-DN (Fig. 5F). Because catalase was localized in the cytosol of peroxisome-deficient RCR-1 cells (Fig. 1, B-D), cytosolic catalase was suggested to be involved in the up-regulation of Bdnf expression. Catalase possesses PTS1-like signal, KANL at the C terminus, whose binding affinity to Pex5p is lower than canonical PTS1 signal, such as SKL (29). We generated RCR-1 cells stably expressing C-terminal ANL-deleted catalase (catalase-⌬C). RCR-1/catalase-⌬C showed the elevation of the Bdnf expression (Fig. 5G). Inactivation of catalase-⌬C by treatment with 20 mM 3AT or mutation of the active site His 75 (30) to Ala (catalase-⌬C-mut) significantly reduced the Bdnf expression (Fig. 5G), implying the requirement of enzymatically active cytosolic catalase. Both catalase-⌬C and catalase-⌬C-mut were mostly localized in the cytosol, but partially detected in organelle fractions (Fig. 5, H and I). Because a catalase tetramer is preassembled in the cytosol prior to peroxisomal import, a part of both catalase-⌬C and catalase-⌬C-mut are likely to form a tetramer with an endogenous catalase monomer, being imported to peroxisomes by a "piggyback" transport mechanism (29,31). Taken together, these results suggested that cytosolic catalase is essential for the elevation of Bdnf expression.
We recently reported that knockdown of PEX5 in a neuroblastoma cell line, SH-SY5Y, induced the up-regulation of

Peroxisome-deficient reductive state abrogates BDNF pathway
BDNF mRNA (10). The treatment with 3AT of siPEX5-transfected SH-SY5Y cells reduced the level of BDNF mRNA to that in control cells (Fig. S6), suggesting that the up-regulation of BDNF in peroxisome-deficient SH-SY5Y cells is induced by the cytosolically mislocalized catalase.

Defect of peroxisome biogenesis induces cytosolic redox state
We earlier reported that the redox state in the cytosol of peroxisome-deficient mutants is more reductive than that of WT cells and that catalase inhibitor, 3AT, induces the cytosolic oxidative state (32), both implying the involvement of cytosolic catalase in the reductive state of peroxisome-deficient mutants. The cytosolic reductive state in peroxisome-deficient mutants is also indicated by increase of the ratio of reduced glutathione (GSH) to oxidized GSH (GSSG). To examine the redox condition of peroxisome biogenesis-defective pex1 CHO ZP107 cells, we determined GSH and other reductive compounds, including NADH and NADPH, by LC coupled with tandem MS (LC-MS/MS). The cell line pex1 ZP107 is defective in peroxi-somal matrix protein import (33). As compared with WT TKa cells, GSH redox index reflecting GSH redox potential, E GSH (34), was relatively elevated in pex1 ZP107 cells (Fig. 6A). NADH/NAD ϩ and NADPH/NADP ϩ levels in ZP107 cells were also higher than those in control cells (Fig. 6, B and C). To investigate the effect of restoration of peroxisome biogenesis on the reductive state in the mutant cell, we assessed the reductive metabolites in pex1 ZP107 cells complemented by stably expressing human PEX1, termed ZP107/PEX1 (33). Levels of GSH redox index, NADH/NAD ϩ , and NADPH/NADP ϩ in ZP107/PEX1 were respectively lowered to the levels in WT TKa cells (Fig. 6, A-C). To assess peroxisomal lipid metabolism, we also determined the levels of plasmalogens and VLCPC in the mutant cells. Plasmalogens were markedly decreased, and VLCPC was accumulated in pex1 ZP107 cells (Fig. 6, D and E), consistent with other PEX-deficient CHO mutant cells, including pex2 Z65 and pex19 ZP119 (35). The defect of peroxisomal lipid metabolism in pex1 ZP107 cells was restored by complementation with PEX1 expression (Fig. 6, D and E). These results  3). H, RCR-1 cells stably expressing FLAG-tagged catalase-⌬C or catalase-⌬C-mut were homogenized and fractionated by ultracentrifugation. Cytosol (S) and organelle (P) fractions were analyzed by SDS-PAGE and immunoblotting with antibodies to FLAG, catalase, Pex14pC, and LDHA. I, RCR-1 cells were treated with 10 g/ml digitonin. Cytosolic catalase activity was determined as in Fig. 1D (n ϭ 3). *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001, by Tukey-Kramer test (A-I). Error bars, S.D.

Peroxisome-deficient reductive state abrogates BDNF pathway
demonstrated that peroxisome biogenesis deficiency induces cytosolic reductive states.

BDNF expression is regulated by cytosolic redox state
We next analyzed the redox states in peroxisome-deficient RCR-1 cells by LC-MS/MS. Similarly to pex1 CHO ZP107 cells, GSH redox index, NADH, and NADPH were relatively increased in RCR-1/Pex5p-DN and RCR-1/Pex14p-DN (Fig. 7,  A and B), confirming that the cellular reductive condition was induced by peroxisome deficiency. Notably, treatment with a catalase inhibitor, 3AT, decreased the reduced forms of metabolites in RCR-1/Pex5p-DN and RCR-1/Pex14p-DN cells (Fig. 5, B-D), as shown above in CHO mutant cells (Fig. 6) (32). Therefore, cytosolic catalase is most likely responsible for the reductive condition in peroxisome-deficient cells. To investigate whether a redox state regulates the Bdnf expression, reductive the state was induced in RCR-1/Venus cells by treatment with n-acetyl cysteine (NAC), a precursor of GSH. Because the treatment with NAC in Neurobasal medium was cytotoxic (data not shown), DMEM/F-12 medium supplemented with 1% FBS was used instead of Neurobasal medium. Bdnf mRNA was indeed up-regulated by the NAC treatment (Fig. 7C). LC-MS/MS analysis revealed that GSH redox index and NADH/NAD ϩ , but not NADPH/NADP ϩ , were elevated by the NAC treatment (Fig. 7,  D and E), consistent with the earlier reports describing the increase of GSH (36) and NADH/NAD ϩ (37), not NADPH/ NADP ϩ (38). Inversely, to induce the oxidative state, the cells were treated with pyruvate that oxidizes NADH to NAD ϩ by the lactate dehydrogenase (LDH) pathway (39,40). Pyruvate treatment indeed gave rise to decrease in NADH/NAD ϩ , thereby reducing the level of BDNF expression in RCR-1/ Pex5p-DN and RCR-1/Pex14p-DN (Fig. 7, F and G). Inhibition of the complex I with rotenone showed no effect on the BDNF mRNA (Fig. 7H), suggesting that the cytosolic NADH/NAD ϩ ratio, not the mitochondrial ratio, regulates BDNF expression. Reducing the level of NADH by treatment with FK866, an inhibitor of nicotinamide phosphoribosyltransferase, induced the down-regulation of Bdnf mRNA (Fig. 7I). Taken together, these results suggest that Bdnf expression is regulated by the cytosolic state of the NADH/NAD ϩ ratio.

Discussion
In this report, we revealed that a defect of peroxisome biogenesis elevates the expression and secretion of BDNF in the glial cell line. Secreted BDNF leads to the axonal branching of primary hippocampal neurons. Up-regulation of BDNF is likely induced by cytosolically mislocalized catalase in a manner dependent on reductive metabolites, including GSH, NADH, and NADPH, in peroxisome-deficient RCR-1 and SH-SY5Y cells. Therefore, these results suggest that the impaired peroxisome biogenesis-dependent cytosolic redox state affects the neuronal morphology via modulating the BDNF expression in glial cells.
Redox state of peroxisome-deficient cells is more reductive than that in the normal cells, such as CHO and RCR-1 cell lines (Figs. 6 and 7), as we earlier demonstrated with a redox state probe, Redoxfluor, using a pex5 CHO mutant, ZP105 (32). This suggests that peroxisomes are also involved in the homeostasis of cellular redox state. Because the redox state becomes oxidative by the exposure of the cells to 3AT, a catalase inhibitor (32), the reductive condition of cytosol in peroxisome-deficient cells, including CHO pex1 mutant and RCR-1 stably expressing dom-

Peroxisome-deficient reductive state abrogates BDNF pathway
inant-negative forms of Pex proteins, is likely owing to the cytosolic catalase. Catalase is a tetrameric peroxisomal matrix enzyme that catalyzes degradation of H 2 O 2 generated by peroxisomal oxidases, including AOx and D-amino acid oxidase (2). The peroxisome-targeting signal of catalase, KANL, possesses a weaker binding affinity to the cytosolic receptor Pex5p than a canonical PTS1 such as SKL. Catalase is mainly localized in peroxisomes, whereas a part of enzymatically active catalase is detected in the cytosol (41)(42)(43). Cytosolic H 2 O 2 is mainly degraded by GSH peroxidase 1 using GSH as a substrate in normal cells (44). In the peroxisome-deficient cells, catalase is localized in the cytosol and degrades H 2 O 2 . Therefore, degradation of H 2 O 2 by cytosolic catalase likely competes with that by GSH peroxidase 1, resulting in the elevation of cytosolic GSH. Because redox ratios of GSH, NADH, and NADPH are closely linked (45), the increase of GSH more likely leads to production of the reduced form of NADH and NADPH. Taken together, the decrease of cytosolic H 2 O 2 caused by mislocalized catalase and the elevation of reductive compounds, including GSH, NADH, and NADPH (Figs. 6 (A-C) and 7 (A and B)), most likely induce the reductive state of the cytosol in peroxisome-deficient cells. Recently, we reported that catalase is released from peroxisomes by BAK-dependent permeabilization of peroxisomal membrane under the oxidative stress condition (43). Oxidative stress also inhibits the catalase import in a manner dependent on Cys 11 of Pex5p (42,46). Cytosolic catalase protects cells from H 2 O 2 -induced oxidative stress (42,43). In addition, catalase is progressively mislocalized to the cytosol as cells become aged and by means of replicative senescence (47,48), where processing of reactive oxygen species are attenuated (49). These findings suggest that the cytosolic redox state is regulated partly, if not completely, via subcellular localization of catalase.
Very recently, we reported that peroxisomal deficiency causes the up-regulation of BDNF and a truncated form of its receptor, TrkB-T1, resulting in malformation of Purkinje cells in the cerebellum of Pex14 ⌬C/⌬C mouse (10). Elevation of BDNF expression is also found in a peroxisome-depleted neuroblastoma cell line, SH-SY5Y, and in the ION neurons of the Pex14 ⌬C/⌬C mouse (10). In this report, we revealed that peroxisome deficiency in astrocytic RCR-1 cells up-regulated the expression and secretion of BDNF, leading to the axonal branching of co-cultured hippocampus neurons. Moreover, in peroxisome-deficient RCR-1 (Fig. 5A) and SH-SY5Y (Fig. S6), the treatment with a catalase inhibitor, 3AT, lowered BDNF expression. These results suggest that the up-regulation of BDNF in the peroxisome-deficient cells is caused by the mislocalized catalase both in a neuroblastoma cell line, SH-SY5Y, and an astrocyte-like cell line, RCR-1. Elevated level of BDNF in neuron was observed in ION of the neonatal Pex14 ⌬C/⌬C mouse, and the BDNF was likely to be delivered to the cerebellum though climbing fibers, resulting in the abnormal morphogenesis of Purkinje cells (10). However, the pathogenic role of BDNF elevation of astrocytes in ZSDs remains defined. Further investigation would delineate how the elevated BDNF affects peroxisome-deficient astrocytes, leading to dysmorphogenesis of CNS.
The Bdnf gene consists of nine exons, a common protein encoded by exon 9 and eight 5Ј noncoding exons (exons 1-8) (50). Transcription of the gene yields BDNF transcripts containing one of the eight 5Ј exons linked to exon 9 or the 5Ј extended coding exon (exon 9A). These splicing variants are

Peroxisome-deficient reductive state abrogates BDNF pathway
thought to have different promoters and be independently regulated (50,51). It remains unknown yet which promoter of the BDNF transcript variant is up-regulated in the peroxisome-deficient cells. The ectopically expressed cytosolic catalase (catalase-⌬C) elevates the Bdnf expression in RCR-1 cells (Fig. 5G). However, the elevated Bdnf level in RCR-1/catalase-⌬C cells is much less than that in RCR-1/Pex5p-DN and RCR-1/ Pex14p-DN (Fig. 3B), thereby raising a possibility that another factor(s) is involved in the elevation of Bdnf transcript. Further investigation is required to address the precise mechanisms underlying the up-regulation of Bdnf expression in peroxisome-deficient cells.
In this report, we demonstrated that the cytosolic reductive states induced by mislocalized catalase elevated the Bdnf expression. Cytosolic catalase is observed in the skin fibroblasts from patients with IRD and less severe ZSDs, whereas punctate structures of PTS1 proteins are discernible (52). These cells showed that abnormalities of peroxisomal metabolisms, such as accumulation of VLCFA and defect of plasmalogen biosynthesis, were only partial (53). LC-MS/MS analysis also revealed a slight reduction of the plasmalogen level but unaltered VLCPC level in the fibroblasts from a patient with IRD. 3 Therefore, a cytosolic reductive state induced by mislocalized catalase is a common phenotype in all cells from patients with ZSDs, including IRD, regardless of their severity. Better understanding of the effect of cytosolic reductive condition on the cellular functions should open a way to the elucidation of pathological mechanisms underlying ZSDs, besides the elevation of BDNF expression.

Primary culture of hippocampus neurons
RCR-1 cells were plated on the paraffin ball-attached coverglass at 4.0 ϫ 10 5 cells/cm 2 . One day after plating, the culture medium was replaced with Neurobasal medium, and the cells were incubated for 2 days.
Primary hippocampal neurons were prepared from Wister rat embryos (embryonic day 18.5 (E18.5)). Briefly, hippocampi were excised into small pieces and dissociated with 15 units of papain (Worthington) in dissociation solutions (0.2 mg/ml L-cysteine, 0.2 mg/ml BSA, and 10 mg/ml glucose) and 0.01% DNase I (Sigma). Cells were separated by gentle trituration passes using a 10-ml pipette and were passed through a 70-m cell strainer (BD Biosciences) to remove large debris. Cells were plated on poly-L-lysine (Sigma)-and laminin (Sigma)-coated plates in Neurobasal medium containing B27 supplement and 0.5 mM L-glutamine. Cell density was 1.0 ϫ 10 5 cells/cm 2 for morphological analysis of axonal development. After 4 h, the culture medium was replaced with the CM derived from RCR-1 cell cultures. RCR-1 cells on a coverglass were overlaid on primary neurons and cultured for 2 DIV (see Fig. 2A). Neuronal morphologies were observed using an inverted Axiovert 200M phase-contrast microscope (Carl Zeiss, Oberkochen, Germany) or AF 6000LX microscope (Leica, Wetzlar, Germany). The length of axons was measured by ImageJ software (National Institutes of Health, Bethesda, MD).
To investigate the potential effect of the RCR-1 cell-derived CM on morphology of primary neurons, RCR-1 cells were plated at a density of 6.5 ϫ 10 4 cells/cm 2 and cultured for 4 days in Neurobasal medium, of which resulting CM was then collected. Primary hippocampal neurons were incubated in the collected CM for 2 DIV. CM containing rBDNF was obtained from RCR-1 cells stably expressing rBDNF, serially diluted with the CM from nontransfected RCR-1 cells, and used for the assay.

Purification of p75ECD-His
A stable transformant of CHO-K1 cells expressing p75ECD-His was cultured in serum-free F-12 medium for 3 days. The cell culture medium was collected and centrifuged to remove floating cells. The resulting supernatant fraction was mixed and incubated with Ni-NTA-agarose beads (Qiagen, Hilden, Germany) for 4 h. The p75ECD-His-bound beads were washed six times with purification buffer (50 mM Hepes-KOH, pH 7.4, 150 mM NaCl, 20 mM imidazole, and 10% glycerol), followed by elution with the purification buffer containing 250 mM imidazole. The eluent was loaded onto PD10 column (GE Healthcare) in suspension buffer (50 mM Hepes-KOH, pH 7.4, 150 mM NaCl, and 10% glycerol) and then concentrated by ultrafiltration in an Amicon Ultra-15 (10,000 molecular weight cutoff; Millipore, Billerica, MA). Purified p75ECD-His was added to the primary neuron cell culture at 1.0 g/ml.

Lipid extraction
Total cellular lipids were extracted by the Bligh and Dyer method (63). Briefly, cells were detached from culture plates by incubation with trypsin and suspended in PBS. Protein concentration was determined by the bicinchonic acid method (Thermo Fisher Scientific). Cell suspensions containing 50 g of total cellular proteins were dissolved in methanol/chloroform/water at 2:1:0.8 (v/v/v), and then 50 pmol of 1-heptadecanoyl-sn-glycero-3-phosphocholine (Avanti Polar Lipids, Alabaster, AL), 1,2-didodecanoyl-sn-glycero-3-phosphocholine (Avanti Polar Lipids), and 1,2-didodecanoyl-sn-glycero-3phosphoethanolamine (Avanti Polar Lipids) were added as internal standards. After incubation for 5 min at room temperature, 1 ml each of water and chloroform was added, and the samples were then centrifuged at 2,000 rpm for 5 min in Himac CF-16RX (Hitachi Koki, Tokyo, Japan) to collect the lower organic phase. To re-extract lipids from the water phase, 1 ml of chloroform was added. The combined organic phase was evaporated under a nitrogen stream, and the extracted lipids were dissolved in methanol.

Extraction of hydrophilic metabolites
Cells were collected in 1 ml of ice-cold methanol and lysed by freeze-thawing two times. Raffinose (50 pmol) and GSH ethyl ester (GSHee, 50 pmol) were added as internal standards. After centrifugation at 21,000 ϫ g for 5 min at 4°C, the supernatant fraction was evaporated at room temperature under nitrogen for 6 h and dissolved in water.

LC-MS/MS
LC-MS/MS analysis of phospholipids was performed as described (35) using a 4000 Q-TRAP quadrupole linear ion trap hybrid mass spectrometer (AB Sciex, Foster City, CA) with an ACQUITY UPLC System (Waters, Milford, MA).
The ⌬E GSH values representing the difference in redox potential of GSH compared with control cells (34) are calculated as follows: ⌬E GSH ϭ ϪRT/2Fln(GSH redox index), where R is the gas constant, T is absolute temperature (K), and F is the Faraday constant.

RT-PCR and real-time RT-PCR
Total RNA was extracted from cells using TRIzol reagent (Invitrogen), and first-strand cDNA was synthesized by a Pri-meScript RT reagent kit (Takara Bio, Shiga, Japan). The expression level of neurotrophin mRNAs was assessed by RT-PCR using the respective sets of primers listed in Table S1. Quantitative real-time RT-PCR was performed with SYBR Premix Ex TaqII (Takara Bio) using an Mx3000P QPCR system (Agilent Technologies, Santa Clara, CA). Several sets of primers used are listed in Table S2.

Immunofluorescent microscopy
Cultured cells were fixed with 4% paraformaldehyde in PBS, pH 7.4, for 15 min at room temperature (64). Peroxisomes were visualized by indirect immunofluorescence staining with the indicated antibodies as described (65). Antigen-antibody complexes were detected with goat anti-mouse and anti-rabbit IgG conjugated to Alexa Fluor 488 and Alexa Fluor 568 (Invitrogen). Phalloidin-TRITC (Sigma) was used for the staining of F-actin. Images were obtained using a laser-scanning confocal microscope (LSM 710 with Axio Observer.Z1; Carl Zeiss).

Immunoblotting
Immunoblotting was performed as described (66). Precision Plus Protein All Blue standards (Bio-Rad) were used as molecular size markers. Immunoblots were developed with ECL prime reagent (GE Healthcare), and immunoreactive bands were detected by X-ray film (GE Healthcare) or an LAS-4000 Mini luminescent image analyzer (Fuji Film, Tokyo, Japan). The band intensities were quantified by Image J software (National Institutes of Health) or Image Gauge software (Fuji Film).

Catalase latency
Catalase latency was evaluated as described (25,43). In brief, trypsinized cells were washed and suspended at 10 6 cells/ml in 0.25 M sucrose and 10 mM Hepes-KOH, pH 7.4. The cells were treated with 10 g/ml digitonin (Wako, Tokyo, Japan) or 1% Triton X-100. After detergent treatment, 20 l of cell suspensions were added to 200 l of H 2 O 2 solution (20 mM imidazole-HCl, pH 7.0, 0.25 M sucrose, 0.1% BSA, and 0.01% H 2 O 2 ) and further incubated for 15 min on ice. After incubation, the catalase reaction was halted by the addition of 200 l Ti(SO 4 ) 2 solution (2.0 M H 2 SO 4 and 1.25% Ti(SO 4 ) 2 ). The concentration of residual H 2 O 2 was determined by absorbance at 410 nm of H 2 TiO 4 using a Benchmark Plus microplate reader (Bio-Rad). Catalase activity was calculated from the decrease in H 2 O 2 concentration after the enzymatic reaction. One unit of activity was defined as the amount of enzyme causing the destruction of 90% of the substrate in 1 min in a volume of 50 ml under assay conditions (41).

Subcellular fractionation and protease protection assay
RCR-1 cells were homogenized with a 27-gauge needle syringe in 500 l of homogenizing buffer (10 mM Tris-HCl, pH 7.4, 0.25 M sucrose, 2 g/ml protease inhibitor mixture (Peptide Institute, Osaka, Japan) including antipain, leupeptin, and aprotinin, and 1 mM phenylmethylsulfonyl fluoride (Nacalai, Kyoto, Japan)) and were centrifuged at 600 ϫ g for 5 min at 4°C. The PNS fraction was ultracentrifuged at 100,000 ϫ g for 30 min to obtain the cytosolic fraction and organelle fraction. Each fraction was analyzed by SDS-PAGE and immunoblotting.
The protease protection assay was performed as described (65,67). Briefly, PNS fractions from each type of RCR-1 cells (4 ϫ 10 5 cells) were treated with 80 g/ml Proteinase K (Sigma) for 30 min on ice in the absence or presence of 0.2% Triton X-100. The reaction was terminated with 1 mM phenylmethylsulfonyl fluoride, and the reaction mixture was separated by ultracentrifugation at 100,000 ϫ g for 30 min at 4°C. Each fraction was analyzed by SDS-PAGE and immunoblotting.

Statistical analysis
Statistical analysis was performed using R software. All Student's t tests used were one-tailed. A p value of Ͻ0.05 was considered statistically significant. Data are shown as means Ϯ S.D. unless otherwise described.

Study approval
The animal ethics committee of Kyushu University approved all animal experiments.

Data availability
All data described are contained within the article and supporting information.