Controlled enzymatic production of astrocytic hydrogen peroxide protects neurons from oxidative stress via an Nrf2-independent pathway

Temporal and Quantitative Regulation of Astrocytic H₂O₂ Production via Rhodotorula gracilis D-Amino Acid Oxidase.

D-amino acid oxidase (DAAO) is a peroxisomal flavoenzyme that oxidatively deaminates D-amino acids into their corresponding imino acids, producing H₂O₂ as a byproduct (20, 21) (Fig. 1A). In contrast to other enzymatic models of H₂O₂ production, such as glucose oxidase, monoamine oxidase, or superoxide dismutase, DAAO permits precise H₂O₂ production via the manipulation of substrate (D-amino acids) availability, which are generally scarce in mammalian cells. We constructed adenoviruses containing flag-tagged cDNA for Rhodotorula gracilis (red yeast) DAAO (rgDAAO) lacking its peroxisomal targeting sequence, which directs rgDAAO expression to the cytoplasm and thus circumvents the avid scavenging of H₂O₂ in peroxisomes by catalase (21). We selected rgDAAO as a H₂O₂ source because it has higher catalytic activity and is less prone to auto-oxidation–induced inactivation than mammalian DAAO (20).

Cytoplasmic rgDAAO staining (via its flag epitope) was observed in ∼80% of primary rat astrocytes transduced with rgDAAO adenoviruses (adDAAO), but was not detectable in empty vector transduced (adEV) astrocytes (Fig. 1B). As expected, the addition of the rgDAAO substrate D-alanine (2 mM, D-Ala) to the bathing media stimulated H₂O₂ production in adDAAO astrocytes (Fig. 1C). Because D-Ala alone was not sufficient to maintain continuous H₂O₂ production beyond 10 min, the bathing media was further supplemented with the DAAO cofactor flavin adenine dinucleotide (FAD) at a range from 0.08 μM to 5 μM (Fig. 1C). It was found that 2.5 μM FAD was required for sustained H₂O₂ production in the presence of D-Ala (Fig. 1C), and thus for subsequent experiments D-Ala treatments were supplemented with 2.5 μM FAD. In the absence of D-Ala, rgDAAO expression did not evoke H₂O₂ production but adding 0.016 to 4 mM D-Ala to the media led to a concentration-dependent increase in the rate of H₂O₂ production (Fig. 1D). We did not detect any increases in H₂O₂ production from control (nontransduced) or adEV-astrocytes with the addition of FAD or any of the D-Ala concentrations tested (Fig. 1D). Thus, H₂O₂ production in cells expressing rgDAAO can be manipulated in a concentration dependent manner by the addition of varying concentrations of D-Ala.

GSH Depletion-Induced Oxidative Stress in Astrocyte-Neuron Cocultures.

Glutamate or its analogs (i.e., homocysteic acid, HCA) induce the neurotoxicity of immature neurons independent of NMDA receptor-mediated excitotoxicity (Fig. S1 A–C) (22). Rather, glutamate inhibits cystine transport and thus leads to the subsequent loss of intracellular cysteine and depletion of the antioxidant GSH, culminating in oxidative stress (22). We sought to define experimental conditions in which the oxidative glutamate toxicity of neurons could be studied using a physiologically relevant ratio of astrocytes to neurons. Thus, primary rat neurons (1 day in vitro, DIV) were plated on top of a confluent monolayer of astrocytes (14–21 DIV) with 5 mM HCA in the bathing medium. Astrocyte-neuron cocultures exposed to 1.25 through 20 mM HCA for 24 h (time when no cell death is observed) showed a concentration-dependent decrease in intracellular GSH levels (Fig. 2A). As expected, exposure to ≥5 mM HCA for 48 h resulted in significant neuronal death (Fig. 2B and Fig. S1D) but no astrocytic death (Fig. 2B and Fig. S1E). To verify a specific effect of GSH depletion in mediating the neuronal death, the bathing media was supplemented with cystine (to overcome competitive inhibition of cystine transport by HCA) and N-acetylcysteine (NAC, an antioxidant that supports GSH synthesis by delivering cysteine independent of the X_c⁻ transporter), both of which partially blocked HCA-induced GSH depletion (Fig. S1F). To verify that the neurons ultimately die because of oxidative stress, vitamin E [a lipophilic antioxidant incapable of preventing GSH-depletion (Fig. S1F)] was added to the cocultures. All of these conditions abrogated HCA-induced neuronal death (Fig. 2C). These studies establish astrocyte-neuron coculture conditions under which neurons can be selectively killed by HCA in the presence of confluent astrocytes.

Astrocytic H₂O₂ Production Evokes Resistant of Neurons to Oxidative Stress.

We next sought to identify kinetic parameters of H₂O₂ production in astrocytes capable of protecting neurons from oxidative stress. To determine the amplitude and duration of astrocytic H₂O₂ production necessary to modulate oxidative neuronal death, we cocultured neurons with adDAAO astrocytes and bathed the cocultures with D-Ala (Fig. 3A). Low-level H₂O₂ production induced by 0.016 mM D-Ala (∼3.7 nmol·min·mg protein) (Fig. 1D) over 7 h protected the neurons from HCA (Fig. 3B). Although low-level H₂O₂ production in adDAAO astrocytes did not affect neuronal viability in the absence of HCA, higher levels of H₂O₂ produced in astrocytes (∼133 nmol·min·mg protein) killed neighboring neurons, even without HCA present (Fig. 3B). D-Ala treatment in control or adEV astrocytes had no effect on neuronal viability (Fig. 3C), and all of the conditions tested did not alter neuronal adherence to the astrocytes (Fig. S2A).

We next used 10 U/mL of the H₂O₂ scavenging-enzyme catalase (CAT), added to the bathing media, to abrogate extracellular adDAAO-generated H₂O₂ released from the astrocytes (Fig. 3D). Indeed, the extracellular catalase prevented neuronal death caused by 2 mM D-Ala–induced H₂O₂ production in the absence of HCA (Fig. 3E). However, the protective effect of low-level astrocytic H₂O₂ production (0.016 mM D-Ala) was not altered by scavenging extracellular H₂O₂ with catalase in HCA-treated cultures (Fig. 3E). These data suggest that high-level H₂O₂ production from astrocytes diffuses out of the cells to exert a neurotoxic effect. In contrast, the neuroprotective effect of low-level astrocytic H₂O₂ production is likely astrocyte-specific. Of note, forced expression of rgDAAO in astrocytes was sufficient to protect mature neurons (14 DIV) from NMDA-induced toxicity independent of D-Ala exposure (Fig. S2 B and C). Such an effect may be the result of long-term endogenous D-serine metabolism by rgDAAO, yielding increased H₂O₂ generation in astrocytes or diminished D-serine (an NMDA coagonist) levels in synaptically active neurons.

To evaluate if DAAO/D-Ala-generated H₂O₂ is working via an intracellular astrocyte-specific mechanism to protect immature neurons, we generated H₂O₂ for 4, 7, or 24 h in adDAAO astrocytes in the absence of neurons (Fig. 4A). We removed the D-Ala from the medium, washed to remove residual H₂O₂ production (Fig. S3A), and added neurons (plus HCA to induce oxidative stress) to the H₂O₂ “primed” astrocytes. Despite never having “seen” H₂O₂, the neurons plated with adDAAO astrocytes (primed for 7 h with 0.016 mM D-Ala to induce low-level H₂O₂ production) were resistant to GSH depletion-induced death (Fig. 4 B and C). The neuroprotective effect was diminished at 24 h and completely absent at 4 h preconditioning (Fig. 4C). There was also a lack of neuroprotection at all time points by high H₂O₂ levels (Fig. 4C) (0.25 and 2 mM D-Ala) that was not a result of astrocyte toxicity (Fig. S3B). Indeed, D-Ala treatment in control or adEV astrocytes had no effect on neuronal viability (Fig. 4D) and all of the conditions tested did not alter neuronal adherence to the astrocytes (Fig. S3C). These studies establish the amplitude and duration-dependence required for astrocytic H₂O₂'s effect on neuronal protection, suggesting that a low level of sustained H₂O₂ production for 7 h (not 4 or 24 h) induces an optimal “state change” in astrocytes that renders them capable of protecting neurons from oxidative death.

To verify that the metabolism of D-Ala by DAAO imparts its effects on astrocytes via H₂O₂, we coexpressed CAT, using an adenoviral vector (adCAT), with adDAAO specifically in astrocytes to scavenge intracellular H₂O₂ production only in astrocytes. Immunoblotting verified increased CAT expression in adCAT-astrocytes and showed that this increased expression persisted in the presence of rgDAAO (Fig. S3D), and the cotransduction of adCAT did not interfere with rgDAAO expression (Fig. S3E). AdDAAO-astrocytes cotransduced with adCAT attenuated H₂O₂ production (Fig. S3F), and this reduction in H₂O₂ by adCAT abrogated the neuroprotective effect of D-Ala in adDAAO astrocytes (Fig. 4E). These results, along with our inability to measure a change in the DAAO-byproduct ammonia at the neuroprotective concentration of D-Ala (0.016 mM D-Ala) (Fig. S3G), support a critical role of H₂O₂ in mediating the astrocyte-dependent neuroprotection.

Nrf2 Activation in Astrocytes Induces Neuroprotection.

Prior studies have suggested that H₂O₂ can activate Nrf2 in astrocytes (18). We therefore examined whether Nrf2, which stimulates the transcription of antioxidant genes and has been implicated in astrocyte-dependent neuroprotection (11, 14), is necessary for the neuroprotective effect of astrocytic H₂O₂. As a positive control for these studies we pretreated astrocytes, before neuronal plating, with the small molecule electrophilic Nrf2 activators sulforaphane (SF), tert-butylhydroquinone (tBHQ) and 15-deoxy-Δ12,14-prostaglandin J2 (15d-PGJ2), or overexpressed Nrf2 using an adenoviral vector. As expected, the electrophiles and Nrf2 overexpression mimicked the neuroprotective effect seen with astrocytic H₂O₂ (Fig. S4 A and B). To verify that the Nrf2 activators induce transcription via the ARE, the binding site for Nrf2, we constructed adenoviruses that express the ARE promoter sequence upstream of a luciferase reporter (ARE-Luc). Indeed, SF activated the ARE in astrocytes (Fig. S4C). To verify that astrocytic Nrf2 is required for the neuroprotection observed with SF, we treated astrocytes with siRNAs targeted against Nrf2 (which decreased Nrf2 protein levels and mRNA levels by 50 to 60%, and abrogated both SF-induced ARE-Luc activity and SF-induced protein levels of HO-1 (a downstream target of Nrf2) (Fig. S4 D–G). Thus, as expected, Nrf2-siRNA blocked the SF-induced astrocytic neuroprotection (Fig. S4H) (which was not a result of siRNA-induced toxicity) (Fig. S4 I and J). In addition, overexpression of the Nrf2 repressor Keap1 abrogated the astrocyte-dependent SF-mediated neuroprotection (Fig S4K).

Astrocytic H₂O₂ Induces Neuroprotection Independent of Nrf2.

With confirmation in hand that Nrf2 is necessary for astrocyte-mediated electrophilic neuroprotection, we next asked whether Nrf2 is also necessary for H₂O₂-induced neuroprotection. Surprisingly, the Nrf2-siRNAs failed to block the neuroprotective effect of adDAAO-mediated astrocytic H₂O₂ production (Fig. 5A). Despite higher levels of neuronal death in cultures exposed to both transfection reagent and HCA, the fold-protection induced by DAAO plus D-Ala treatment in the presence of transfection reagent and Nrf2-siRNAs remained consistent with the previously observed protection (compare Fig. 4C with Fig. 5A). We believe it unlikely that the persistent neuroprotection reflects the incomplete knockdown of Nrf2, because adDAAO-induced H₂O₂ production failed to augment markers of Nrf2 activation, such as HO-1 protein levels (Fig. 5B) and ARE-Luc activity (Fig. 5C). We also confirmed that the lack of effect of adDAAO-mediated H₂O₂ production on ARE-Luc activity was not a result of diminished rgDAAO expression (Fig. S5A), diminished adARE-Luc transduction (Fig. S5B), or a lack of H₂O₂ production (Fig. S5C) in astrocytes cotransduced with adDAAO and adARE-Luc. We also confirmed that exogenous treatment of astrocytes with H₂O₂ (up to 30 μM H₂O₂) failed to initiate significant ARE-Luc activity (Fig. 5D). We were unable to test higher levels of exogenous H₂O₂ on ARE-Luc activity because of astrocyte toxicity (Fig. S5D). Finally, the Nrf2 electrophilic activators SF or tBHQ maintain their astrocytic protective effect even with overexpression of the H₂O₂ scavenging enzyme CAT, further supporting disparate mechanisms of the electrophile and H₂O₂ protective pathways (Fig S5E).

Because Nrf2-dependent neuroprotection correlates strongly with enhanced astrocytic GSH release (11), we examined if the astrocyte-dependent neuroprotective effect could be sustained with astrocyte conditioned media, and if astrocytic GSH levels increase in response to adDAAO-generated H₂O₂. Consistent with a lack of Nrf2 involvement in the H₂O₂ effect, conditioned media from SF-treated astrocytes protected the neurons from oxidative stress, but the media from H₂O₂-primed astrocytes failed to induce neuroprotection (Fig. S6A). In addition, SF, but not adDAAO-derived H₂O₂, enhanced astrocytic intracellular GSH levels (Fig. S6B). Finally 7-h D-Ala pretreatment in adDAAO astrocytes before neuronal plating with HCA failed to preserve GSH levels in the cocultures (Fig. S6C).

Microarray Analysis of Astrocytic Low-Level DAAO-induced H₂O₂ Generation.

As other prosurvival transcription factors, such as NF-κB, are also redox-regulated, we performed an unbiased analysis of the transcriptional changes induced by low-level H₂O₂ (induced by 0.016 mM D-Ala) in adDAAO astrocytes using two microarray approaches: (i) high density genome-wide and (ii) custom-designed microarray platform. The genome-wide approach showed no significant transcriptional changes, including no changes in a battery of Nrf2-regulated genes, in response to low-level H₂O₂ generation (Figs. S6D and S7). However, the more sensitive custom-designed microarray approach uncovered several significant changes induced by low-level H₂O₂, most notably the up-regulation of interleukin 1-β (IL-1β) and caveolin 1 (CAV1), and the down-regulation of transglutaminase 2, an enzyme whose selective inhibition has been associated with neuroprotection (23) (Table S1). The H₂O₂-induced regulated genes were almost entirely independent from genes controlled by the Nrf2-activator SF, such as GFAP (Tables S2 and S3). Together, these microarray data further support the lack of involvement of Nrf2 in mediating the H₂O₂ effect and provide insight into potential low level H₂O₂-dependent mechanisms.

PTP Inhibition Mimicks the Neuroprotective Effect of Astrocytic H₂O₂.

PTPs are uniquely and highly sensitive to oxidation-induced inactivation by physiological H₂O₂ levels because of the low pK_a of cysteine residues found within their catalytic domain (2). Indeed, PTP inhibition can enhance interleukin levels (24), and IL-1β was up-regulated in our gene-array analysis (Table S1). Thus, to establish if low-level H₂O₂-induced PTP inactivation is a plausible mechanism responsible for the astrocyte-dependent neuroprotective effect, astrocytes were pretreated with three structurally diverse PTP inhibitors before neuronal plating. Indeed, sodium orthovanadate, phenylarsine oxide, and BVT-948 mimicked the astrocytic H₂O₂-induced neuroprotection (Fig. 6 A–C and Fig S8 A–C). Consistent with the H₂O₂-induced neuroprotection, BVT-948 (which inhibits PTPs via H₂O₂ generation) was most effective at the lowest concentration tested.

Cell Culture.

Primary astrocyte cultures were prepared from the cerebral cortices of rat pups (P1–3). Primary neuronal cultures were prepared from the forebrains of rat embryos (E17). Dissociation was performed as described by Ratan et al. (22).

Adenoviral Transduction.

Cultured astrocytes were treated with the adenoviral vectors at a multiplicity of infection of 15 for 4 h in serum-free medium.

Extracellular H₂O₂ Measurement.

A HRP/Amplex Red substrate system was used to monitor extracellular H₂O₂ levels in live astrocytes.

Immunocytochemistry.

Cells were paraformaldehyde-fixed and incubated with primary/secondary antibodies and mounted with a DAPI-containing solution.

GSH Measurements.

Intracellular total GSH was determined using the GSH-Glo(TM) Glutathione Assay kit (Promega).

Neuronal Viability.

Fixed cocultures were incubated with antibodies for the neuronal specific marker MAP2 followed by HRP-conjugated secondary antibodies. HRP activity was then monitored using an Amplex Red reaction buffer.

Astrocyte Viability.

Astrocyte viability was quantified using the CellTiter 96 Assay (Promega).

Live-Cell Imaging.

Phase contrast images were taken with a Nikon Eclipse TS 100.

Western Blots.

Following SDS/PAGE, transfer to nitrocellulose membrane and primary/secondary antibody incubation, astrocyte protein bands were detected with an infrared imaging system (LI-COR Biosciences).

ARE-Luciferase Reporter Assay.

Astrocytes were transduced with adenoviral vectors containing the ARE sequence upstream of a luciferase reporter. Luciferase activity was measured using a luciferase assay kit (Promega).

Small Interfering RNA Transfections.

Nrf2 knock down was achieved using siRNA targeted against Nrf2. Cultured astrocytes were transfected with the siRNAs plus the transfection reagent TransIT-siQUEST (Mirus) overnight.

Quantitative PCR.

Total RNA was isolated from primary rat astrocytes using the NucleoSpin RNA II kit. A TaqMan RNA-to-CR_TR 1-Step Kit was used for real-time PCR.

Statistical Analysis.

Unless otherwise stated, statistical significance was determined by a one-way ANOVA followed by posthoc Newman Keuls or Dunnett's test using Statistica software (StatSoft). To determine the significance of the rate of H₂O₂ production, a main-effects ANOVA was performed.