Mild oxidative stress activates Nrf2 in astrocytes, which contributes to neuroprotective ischemic preconditioning

Haskew-Layton et al. (1) reported that subtoxic doses of H₂O₂ fails to activate nuclear factor erythroid 2-related factor (Nrf2) in astrocytes and triggers Nrf2-independent responses that protect cocultured neurons. Contrary to this, we show that mild oxidative insults, including subtoxic H₂O₂, strongly activate astrocytic Nrf2/antioxidant response element (ARE)-dependent gene expression, which, moreover, contributes to neuroprotective ischemic preconditioning.

In mixed neuron/astrocyte cultures (2, 3), treatment with physiologically relevant H₂O₂ doses (25–100 μM, similar/less than those recorded postischemia) (4) induced Nrf2-target genes sulfiredoxin (Srxn1) and heme-oxygenase 1 (Hmox1) in wild-type but not Nrf2^−/− cultures (Fig. 1). Similarly, exposing cultures to oxygen–glucose deprivation (OGD; an in vitro ischemia model), followed by reoxygenation, also induced Nrf2-target genes. Induction of Hmox1 in mixed cultures was restricted to astrocytes (Fig. 1), and Nrf2-target gene induction was not observed in enriched neuronal cultures (<0.2% astrocytes) (Fig. 1), strongly suggesting that astrocytes are the sole locus for Nrf2 activation by oxidative stress. Furthermore, study of enriched Nrf2^+/+ and Nrf2^−/− astrocyte cultures showed clear H₂O₂ (and OGD)-induced Nrf2-dependent gene activation, contrary to that reported in ref. 1.

Fig. 1.

Subtoxic oxidative insults activate Nrf2-dependent gene expression in astrocytes. (Upper and Lower Left) Cells were treated with H₂O₂ or 5 μM tert-Butylhydroquinone (tBHQ) for 6 h or subjected to a 3-h episode of oxygen–glucose deprivation (OGD), after which they were returned to normoxia for an additional 3 h. Expression of Srxn1 and Hmox1 mRNA was analyzed by quantitative real-time PCR as described (3) and normalized to Gapdh. Cell types treated include WT and Nrf2^−/− mixed neuronal/astrocyte cultures [90% neuronal nuclear protein (NeuN⁺) neurons and 10% GFAP⁺ astrocytes, made as described] (2, 3) as well as highly enriched neuronal cultures (>98% NeuN⁺ neurons and <0.2% GFAP⁺ astrocytes) obtained by preventing astrocytic proliferation by treating cultures with the antimitotic cytosine-arabinoside immediately postplating. Astrocyte cultures (Lower Left; >96% GFAP⁺) were obtained by plating at low density in DMEM + 10% FBS. All cell types were used after 9–11 d in culture. For OGD, cells were placed in degassed medium containing mannitol instead of glucose and put in a modular incubator chamber flushed with 95% N₂ and 5% CO₂ according to manufacturer's instructions (Billups-Rothenberg, Del Mar, CA). *P < 0.05 Bonferonni t test (n = 4–8). (Inset) Concurrent immunofluorescent detection of Gfap and Hmox1 protein in mixed neuronal/astrocyte cultures treated as indicated for 24 h. Example pictures illustrate the localization of Hmox1 expression to GFAP+ astrocytes. GFAP monoclonal antibody clone GA5 (Sigma) and Hmox1 polyclonal antibody (Stressgen). (Lower Right) H₂O₂ and OGD treatments that induce Nrf2-dependent gene expression in astrocytes are not toxic. Astrocytes were treated with the indicated concentrations of H₂O₂ for 24 h or subjected to 3 h OGD followed by 21 h in normoxia and normal medium. Cellular viability was assessed using the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, WI) and also by assessing nuclear integrity of random fields of cells (>500 cells counted per treatment across three independent experiments; levels of nuclear pyknosis over and above control levels are shown). The CellTiter-Glo Luminescent Cell Viability Assay measures cellular ATP levels, and for each experiment, a standard curve of ATP concentrations was included to ensure that the levels of signal obtained from the cell samples were within the linear range of the assay.

One possible explanation for this discrepancy lies in their Nrf2 assay: a luciferase reporter incorporating the ARE of the NQO1 promoter (1). Different AREs can have different Nrf2 dependencies for basal and/or inducible activity, and we observe relatively weak Nqo1 induction by 100 μM H₂O₂ (2.1- ± 0.06-fold; n = 5). Basal Nrf2 activity seems sufficient for strong Nqo1 expression in astrocytes: basal Nqo1 expression in Nrf2^−/− cultures is only 14 ± 2% of that in WT. Another potential explanation is that H₂O₂ doses >30 μM were not studied, because 100 μM were reportedly toxic based on 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium (MTT) assay (1). However, the MTT assay may overstate toxicity, because it measures cellular NAD(P)H-dependent reducing activity, which could drop if subtoxic stress causes NAD(P)H levels to fall. We see no evidence of H₂O₂ toxicity in astrocytes up to 200 μM, as assessed by either ATP assay or nuclear integrity (Fig. 1).

In contrast to the reported Nrf2-independence of adaptive astrocytic neuroprotective responses acting on immature neurons (48 h in culture) (1), we find that astrocytic Nrf2 contributes to adaptive neuroprotective responses in more mature neurons (Fig. 2). A brief, nonneurotoxic episode of OGD (90 min) applied to mixed cultures preconditions neurons against a subsequent neurotoxic OGD episode 24 h later (Fig. 2). This preconditioning episode activates Nrf2 in mixed cultures (Srxn1: 1.97- ± 0.06-fold; Hmox1: 1.48- ± 0.08-fold; n = 6) but not in pure neuronal cultures, and Hmox1 induction is restricted to astrocytes. In Nrf2^−/− cultures, neuronal vulnerability to OGD was similar to Nrf2^+/+ cultures (Fig. 2). However, the brief OGD-induced preconditioning effect was substantially lower in Nrf2^−/− cultures (Fig. 2), strongly implicating Nrf2 activation in ischemic preconditioning. This response may also be relevant in vivo: a standard preconditioning inducing stimulus in adult mice (15-min occlusion of the middle cerebral artery) triggered Nrf2-target gene induction in the ipsilateral cortical hemisphere (Fig. 2). Finally, we observe that subtoxic H₂O₂ also induces Hmox1/Srxn1 expression in human ES cell-derived astrocytes (Fig. 2), suggesting that human Nrf2 is activated by mild oxidative stress. Thus, in addition to Nrf2-independent pathways (1), astrocytic Nrf2-dependent pathways are likely to be important mediators of neuroprotective adaptive responses to oxidative stress.

Fig. 2.

Nrf2 contributes to the protective effect of preconditioning. (Upper Left) A subtoxic episode of OGD protects neurons against a subsequent toxic insult. Preconditioned (PC) mixed cultures (Nrf2^+/+ and Nrf2^−/−) were subjected to 1.5 h of OGD, 24 h before a 3-h OGD exposure, after which they were returned to normal conditions for a further 21 h. Thereafter, cells were fixed, and death was assessed by blind counting of nuclear pyknosis (1,500–3,000 cells per treatment). *P < 0.01 (ANOVA followed by Tukey's posthoc test; n = 11 Nrf2^+/+; n = 5 Nrf2^−/−). Note that the 3-h OGD insult is not toxic to the astrocytes in the culture (Fig. 1). Although preconditioning is observed in both Nrf2^+/+ and Nrf2^−/− cultures, the degree of protection is less in the Nrf2^−/− cultures. (Upper Right) The effect of preconditioning is lower in Nrf2^−/− cultures. The effect of preconditioning was calculated as the percentage of OGD-induced death that was prevented by the prior preconditioning episode. *P < 0.01 (Student t test; n = 11 Nrf2^+/+; n = 5 Nrf2^−/−). (Lower Left) Transient ischemia activates Nrf2 target genes in vivo. Srxn1 and Hmox1 expression was assessed 4 h after mice were subjected to a 15-min transient occlusion of the middle cerebral artery (MCAO), a stimulus known to elicit preconditioning in vivo. *P < 0.01 (paired t test; n = 6). The MCAO procedure was performed exactly as described (2), after which the mice were killed by terminal anesthesia and brains were snap-frozen in liquid nitrogen. RNA was then extracted from both ipsi- and contralateral cortical hemispheres and subjected to qPCR analysis of Srxn1 and Hmox1 expression, normalized to Gapdh levels. (Lower Right) Sub-toxic doses of H₂O₂ induce Nrf2 target gene expression in human embryonic stem cell-derived astrocytes. Astrocytes were derived from the Hues9 human embryonic stem cell (hESC) line as described (5). After 7 d of differentiation, hESC-derived astrocytes were treated with 50 μM H₂O₂, RNA was harvested 6 h later, and levels of SRXN1 and HMOX1 were analyzed by quantitative realtime-PCR (normalized to GAPDH levels). *P < 0.05 (n = 3).