Overwhelming of the thioredoxin system by oxidative trauma

Of the 2-Cys Prxs, mammalian neurons predominantly express PrxII, III and IV¹⁹. Consistent with this, a pan-2-Cys Prx antibody detected 3 bands (Fig. S2a). To identify the bands, we over-expressed PrxII, III and IV in turn and found enrichment of the bottom, middle and upper endogenous band repectively (Fig. S2a), consistent with their theoretical molecular weights.

The existence of overoxidized Prx-SO_2/3H indicates an overwhelmed thioredoxin-peroxiredoxin system. Western analysis using a Prx-SO_2/3H-specific antibody revealed that H₂O₂ caused Prx overoxidation in control and MK-801-treated neurons, while BiC/4-AP-treated neurons displayed far less overoxidation (Fig. 2a). MK-801 decreased the effect of BiC/4-AP (Fig. S2b). These differences were not due to BiC/4-AP-induced expression of core components of the thioredoxin-peroxiredoxin system: levels of thioredoxin, thioredoxin reductase and 2-Cys Prxs were unchanged by BiC/4-AP, or by MK-801 (Fig. 2b,c). To find an alternative explanation, we performed microarray expression analysis on neurons experiencing different levels of synaptic NMDAR activity. We searched for genes that were regulated in opposite directions by enhancing/suppressing synaptic NMDAR activity in vitro (BiC/4-AP vs. MK-801 treatment), and similarly regulated by NMDAR blockade in vivo. We used mouse arrays, since they offered better annotation, and mouse cortical neurons exhibited similar activity-dependent resistance to oxidative stress (data not shown).

Open in a separate window

Fig. 2

Synaptic activity prevents the overoxidation of peroxiredoxins in response to an oxidative insult, and negatively regulates the Thioredoxin inhibitor Txnip

A) Western analysis of Prx overoxidation using an anti-PrxSO_2/3H specific antibody. Analysis normalized to appropriate Prx band intensity. *p<0.05 compared to control, H₂O₂-treated neurons (n=5). Note for upper (PrxIV) band, a higher exposure is often taken to more accurately assess loading (as is the case in the example shown). B,C) Thioredoxin, Thioredoxin reductase and 2-Cys Prx levels analyzed by Western blot after 24 h of the indicated treatments (n=3). D) q-RT-PCR of Txnip RNA from rat cortical neurons treated as indicated (this and all subsequent q-RT-PCR data are normalized to Gapdh levels. *p<0.05 compared to control (n=4)). E) Western analysis of Txnip protein expression in response to the indicated treatments (24 h, *p<0.05 compared to control, n=3). This and all quantitation of Western blots involves normalisation to β-tubulin or other suitable control (such as Akt in the case of Pi-Akt, Prx in the case of PrxSO_2/3H). F) Co-immunoprecipitation of Thioredoxin with an anti-Txnip antibody in samples from neurons experiencing different levels of synaptic NMDAR activity for 24 h (*p<0.05, n=5). G) Thioredoxin activity (insulin-reducing assay) in neurons treated with MK-801, expressed as a % of the activity observed in BiC/4-AP-treated neurons (*p<0.05 compared to BiC-stimulated neurons, n=6).

Pro-oxidative Txnip is suppressed by synaptic activity

Seven genes were induced by BiC/4-AP treatment (in a MK-801-sensitive manner) which were also suppressed by MK-801 in vitro and in vivo (supplemental table T1). One gene exhibited negative regulation by synaptic NMDAR activity: Txnip (thioredoxin interacting protein) which binds thioredoxin, inhibits its activity and promotes vulnerability to oxidative stress^{3, 20}, and was prioritized for further study. We confirmed NMDAR-dependent Txnip regulation by q-RT-PCR in mouse (not shown) and rat cortical neurons (Fig. 2d) and by western blot (Fig. 2e). A timecourse revealed Txnip expression to be dynamically regulated: both rapidly suppressed by activity, and quickly up-regulated following activity blockade with TTX (Fig. S2c). MK-801 and TTX combined did not induce Txnip more than either drug alone (Fig. S2d), consistent with TTX acting by preventing spontaneous synaptic NMDAR activity. Co-immunoprecipitation studies revealed that in MK-801-treated neurons (high Txnip levels), Txnip is bound to thioredoxin (Fig. 2f) which would be expected to inhibit thioredoxin activity. We measured thioredoxin enzyme activity in cell extracts taken under low detergent conditions designed to preserve Txnip-thioredoxin interactions. Enzyme activity in our cultures is almost entirely neuronal: glial cells contribute only 1-2.5 % of thioredoxin activity (see supplemental methods). Consistent with an inhibitory effect of Txnip, extracts from MK-801-treated neurons displayed lower levels of thioredoxin activity than extracts from BiC/4-AP-treated neurons (Fig. 2g), despite equivalent thioredoxin protein levels at 24 and 48 h (Fig. 2b and data not shown). To see whether Txnip renders neurons vulnerable to oxidative stress we over-expressed it in synaptically active neurons, with an eGFP marker to track neuronal fate. >99% of eGFP-expressing cells were NeuN-positive, and <1% were GFAP-positive (data not shown). Expression of Txnip did not alter BiC/4-AP induced bursting (Fig. S3a). By monitoring the neurons before and after H₂O₂ treatment, we found that Txnip expression reduced the antioxidative effects of synaptic activity (Fig. 3a). Consistent with its known function, thioredoxin overexpression was neuroprotective against a H₂O₂ insult (Fig. 3a). We corroborated the effect of Txnip using a different method of monitoring the viability of transfected neurons: co-expression of a constitutively active luciferase expression construct (Fig. 3b). Thus, Txnip influences neuronal antioxidant defences.

Open in a separate window

Fig. 3

Txnip and Thioredoxin can regulate neuronal vulnerability to oxidative stress

A) Loss of GFP-positive neurons expressing Txnip, Thioredoxin (Txn) or control plasmid (beta-globin) in the face of 24 h 100 μM H₂O₂. *p<0.05 (n=4). B) Neurons transfected with a constitutively active vector (SV40-Luc) and Txnip or control plasmid, treated with BiC/4-AP prior to oxidative insult for 24 h, followed by luciferase activity measurement *p<0.05 (n=5). C(left) q-RT-PCR of Txnip mRNA expression in the murine cortex in vivo in response to MK-801 treatment at P0 and P7. *p<0.05 (n=4). C(right) Western blot showing up-regulation of Txnip protein in the cerebral cortex of individual mice in vivo by MK-801 treatment of P7 mice. Each lane represents a different mouse. D) RNA from post-mortem human frontal cortices was extracted and TXNIP mRNA abundance assessed by quantitative RT-PCR. TXNIP is shown normalized to GAPDH and 28S, as their expression in the frontal cortex does not change significantly with age⁴⁹(n=16). E) Neurons transfected with rTxnip plus control or one of two Txnip-directed siRNAs. Protein harvested at 24 h and subjected to Western analysis for Txnip protein. Untransfected sample (U/T) shown for comparison. F) Loss of GFP-positive neurons transfected with control or one of two Txnip siRNAs in the face of 24 h 100 μM H₂O_2. Scale bar 40 μm. Data from 4 independent experiments.

Cortical neuronal death induced by NMDAR blockade is developmentally regulated: at P0 none is observed but by P7 cell-death is widespread^{13, 21}. We found that the coupling of NMDAR blockade to Txnip up-regulation in vivo is similarly regulated. At P0, MK-801 does not induce an increase in Txnip, in contrast to P7 (Fig. 3c), suggesting that Txnip may contribute to cell death at P7. The absence of an effect at P0 could be due to lower NMDAR expression at this stage. We also investigated whether Txnip mRNA expression varies with age in humans, and found elevated levels in the cortices from old individuals (81-95 yr) compared to young (25-37 yr, Fig. 3d).

To determine whether suppressing Txnip expression is neuroprotective we investigated the effect of siRNA-mediated knockdown. Validation of Txnip-directed siRNA was hampered by the absence of a Txnip antibody which detects endogenous neuronal Txnip by immunohistochemistry, and the fact that at this developmental stage siRNA transfection efficiency is too low to analyze knock-down of endogenous protein by Western blot. Since Txnip protein is sufficiently unstable that suppression of transcription is reflected at the protein level after 24 h (Fig. 2e), we can reasonably assume that effective siRNA is able to knock down endogenous protein levels in a similar timeframe. We expressed rat Txnip in neurons and at 24 h found that co-transfection of two different Txnip-, but not control-siRNA, blocked Txnip expression (Fig. 3e). Neither rat Txnip siRNA impaired expression of transfected human Txnip (Fig. S3c) which contains 2 and 4 nucleotide differences (in central regions) respectively from the corresponding rat sequence.

We found that Txnip siRNA did not significantly protect H₂O₂-exposed unstimulated neurons (Fig. 3f) indicating, not surprisingly, that multiple transcriptional changes are needed to mimic the antioxidative effect of synaptic NMDAR activity.

Induction of Prx-SO_2/3H-reducing genes Sesn2 and Srxn1

To see whether overoxidized Prxs could be reduced in neurons, we applied a brief, high (200 μM) dose of H₂O₂ to induce Prx overoxidation and looked for recovery after H₂O₂ washout. We focussed on PrxII, the major neuronal cytosolic Prx¹⁹. Both unstimulated and MK-801-treated neurons exhibited no reduction in Prx-SO_2/3H 16 h after washout (Fig. 4a). In contrast, BiC/4-AP-treated neurons exhibited a large reduction in post-washout Prx-SO_2/3H levels (Fig. 4a). Thus, the degree of synaptic activity experienced can determine whether Prx overoxidation is reversible, which may affect the efficacy of H₂O₂ detoxification by the thioredoxin-peroxiredoxin system.

Open in a separate window

Fig. 4

Synaptic activity promotes reduction of Prx-SO_2/3H and induces neuroprotective expression of the Prx-SO_2/3H-reducing genes Sesn2 and Srxn1

A) Recovery of PrxII-SO_2/3H following a 1 h exposure to high H₂O₂ (200 μM). Neurons stimulated as indicated for 12 h prior to H₂O₂ exposure. Washout period was 16 h (*p<0.05 compared to pre-washout level, n=5). B,C) q-RT-PCR of Sestrin2 (Sesn2, B) and Sulfiredoxin (Srxn1, C) in response to the indicated treatments (BiC stimulation: 4 h). *p<0.01 (n=7). Lower panels show regulation of Sesn2 and Srxn1 protein. D) H₂O₂-induced loss of GFP-positive neurons transfected as indicated. Scale bar 60 μm.*p<0.05 (one-way ANOVA followed by Fisher’s LSD post-hoc test for this and subsequent siRNA experiments, n=4). Right panels show example pictures (arrows indicate transfected neurons whose fate is studied, Scale bar 60 μm). E) Effect of control- or Sesn2-directed siRNA on BiC/4-AP (24 h) induced Sesn2 expression. F) Effect of control- or Srxn1-directed siRNA on production of rat Srxn1 driven from an expression vector. G) Effect of knock-down of Sesn2 and Srxn1 on activity-dependent protection against an oxidative insult. Neurons were transfected with control siRNA or one of two pairs of siRNAs which target both Sesn2 and Srxn1. Neurons were stimulated where indicated with BiC/4-AP and then treated with 100 μM H₂O_2.. *p<0.05 unpaired T-test (n=4). H) Effect of knock-down of Sesn2 and Srxn1 on long-lasting activity-dependent protection. Experimental details as for (G) except that BiC/4-AP-induced synaptic activity was terminated after 12 h by the addition of TTX + MK-801. After this, the oxidative insult (100 μM H₂O₂) was applied. *p<0.05 (n=6).

We next studied the transcriptional regulation of the two genes whose products mediate reduction of Prx-SO_2/3H: sulfiredoxin (Srxn1,⁸) and sestrin2 (Sesn2​⁹). BiC/4-AP strongly induced Srxn1 and Sesn2 mRNA and protein expression (blocked by MK-801, Fig. 4b,c), offering an explanation for the activity-dependent enhancement of Prx-SO_2/3H-reducing capacity (Fig. 4a). We hypothesized that up-regulation of Sesn2 and Srxn1 may cooperate with the activity-dependent down-regulation of Txnip in promoting resistance to oxidative insults. Overexpression of Sesn2 and Srxn1 in a cell-line (HEK293) that is amenable to high efficiency transfection revealed that, following H₂O₂ treatment, Sesn2/Srxn1-coexpressing cells have less Prx-SO_2/3H than control cells (Fig. S4), in agreement with previous studies^{9, 22, 23}. We then mimicked the three activity-dependent events described above by transfecting expression vectors for Sesn2 and Srxn1 as well as Txnip siRNA. We found a reduction in cell death following a H₂O₂ insult (Fig. 4d). Expression of Sesn2 and Srxn1 alone afforded variable protection that did not reach significance.

We next performed experiments using siRNA targeted against Sesn2 and Srxn1. Activity-dependent induction of Sesn2 expression was visualized by immunofluorescence, and two siRNAs targeted against Sesn2 (siRNA(1) and (2)) knocked down expression (Fig. 4e). Validation of Srxn1-directed siRNA was hampered by the absence of a Srxn1 antibody which detects endogenous neuronal Srxn1 by immunohistochemistry. Instead, we overexpressed Srxn1 plus a eGFP marker, and analyzed the effectiveness of the siRNA in preventing Srxn1 expression, finding two which worked: siRNA(3) and (4), (Fig. 4f).

We paired the siRNAs (1/3 and 2/4) to test the effect of combined knockdown of Sesn2 and Srxn1. BiC/4-AP-induced protection was diminished in neurons transfected with either pair of siRNAs (Fig. 4g). We also tested the effect of these siRNA pairs on long-lasting protection afforded by a prior episode of synaptic activity (Fig. 1f) and found a similar effect (Fig. 4h). These experiments indicate that the combined changes in Txnip, Sesn2 and Srxn1 expression contribute to the enhancement of intrinsic antioxidant defences mediated by synaptic NMDAR activity.

Txnip is a FOXO target gene

Since Txnip, Sesn2 and Srxn1 are novel activity-regulated genes, we wanted to understand how they are regulated. Examination of the Txnip promoter revealed a conserved consensus site for the Forkhead box O (FOXO) class of transcription factors (Fig. 5a). FOXOs undergo nuclear export in response to Akt-dependent phosphorylation²⁴ and Akt is activated by synaptic NMDAR activity via phosphatidylinositol-3-kinase (PI3K​¹⁸). Consistent with a role for FOXOs, inhibition of PI3K with LY294002 elevated basal Txnip mRNA and protein expression, and blocked activity-dependent suppression of Txnip (Fig. 5b). We next cloned the Txnip promoter upstream of a luciferase reporter (Txnip-Luc) and introduced a mutation into the putative FOXO binding site (Txnip(mut)-Luc). Txnip-Luc exhibited the expected activity-dependent suppression (Fig. 5c), while Txnip(mut)-Luc exhibited lower basal activity and did not undergo activity-dependent suppression (Fig. 5c). Overexpression of the major neuronal FOXOs, FOXO1 or FOXO3a, enhanced Txnip-Luc promoter activity but failed to activate Txnip(mut)-Luc (Fig. 5d). Thus, Txnip contains a functional FOXO site which mediates activity-dependent inactivation of the promoter. Moreover, we found that BiC/4-AP treatment triggered PI3K-dependent FOXO1 phosphorylation (Fig. 5e) and nuclear export (Fig. 5f). In contrast, MK-801 enhanced nuclear localisation of FOXO1 (Fig. 5f). A mutant FOXO with its Akt phosphorylation sites mutated (FOXO-ADA​²⁵) did not undergo activity-dependent export (data not shown). Chromatin-IP experiments revealed that endogenous FOXO1 associates with the Txnip promoter in MK-801-treated neurons more strongly than in BiC/4-AP-treated neurons (Fig. 5g,h). As a negative control, chromatin-IP with a phospho-FOXO1 antibody failed to immunoprecipitate the Txnip promoter (Fig. 5h). Consistent with the intermediate nature of the untreated neurons with regard to Txnip expression and FOXO localisation, they exhibited variable FOXO promoter occupancy: sometimes resembling MK-801-treated neurons and sometimes BiC/4-AP-treated neurons (Fig. 5h left vs. right). These data support a model whereby synaptic activity turns off Txnip ranscription by inducing the PI3K-Akt pathway, triggering FOXO phosphorylation, dissociation from the Txnip promoter and export from the nucleus.

Open in a separate window

Fig. 5

Txnip is a FOXO target gene

A) Schematic showing conservation of a FOXO binding site in the proximal 5′ promoter region of Txnip, positions given relative to start of protein coding region (NB. 5′UTR is 222 nt long). B) PI3K inhibition by LY294002 (30 μM) induces Txnip expression, and suppression of Txnip protein and mRNA expression by BiC treatment is blocked by LY294002 *p<0.05 compared to control (n=3). C) Txnip-Luc and Txnip-(mut)-Luc (FOXO site mutated) activity assayed 16 h after the indicated treatments (n=5). In all reporter assays, firefly luciferase based reporter signal is normalized to expression of a cotransfected renilla luciferase control plasmid, pTK-RL. D) Txnip-Luc and Txnip-(mut)-Luc activity assayed in neurons co-transfected with Control or FOXO expression plasmids. *p<0.05 compared to control plasmid (n=3-5). E) Western analysis of FOXO1 phosphorylation in response to the indicated treatments (30 min stimulation, example Western shown below analysis). *p<0.05 compared to control (n=3). F) Subcellular distribution of transfected myc-tagged FOXO1 analyzed by immunofluorescence (examples on the right; scale bar 30 μm.). G,H) Chromatin immunoprecipitation with an anti-FOXO1 antibody, followed by PCR of the Txnip promoter region containing the consensus FOXO binding site (compared to PCR of the input). (G) shows analysis of ChIP band intensity (Normalized to input) relative to MK-801-treated neurons *p<0.05 (n=5). (H) shows example individual ChIP experiments, with (H, right) showing an anti-phospho-FOXO1 negative control ChIP.

Sesn2 is a C/EBP target gene

We created a Sesn2-Luciferase reporter, which was activated by BiC/4-AP-induced NMDAR activity (Fig. S5a). Deletion analysis identified the activity-inducible region between -494 and -107 relative to the transcription start site (Fig. 6a). Further deletions revealed that inducibility was conferred by 2 regions: -378 to -249 and -249 to -107 (data not shown). In silico analysis revealed a potential binding site for the transcription factor CCAAT enhancer binding protein (C/EBP) in each of these regions. Moreover, expression of an interfering mutant of C/EBP, ΔC/EBP²⁶, inhibited activity-dependent induction of Sesn2-Luc (Fig. 6b). Mutation of the proximal putative C/EBP site impaired activity-dependent induction of Sesn2-Luc (Fig. 6c), while mutation of both sites abolished induction (Fig. 6c). Thus, induction of Sesn2 by synaptic activity is mediated largely by C/EBP.

Open in a separate window

Fig. 6

Sesn2 is a C/EBP target gene, Srxn1 is an AP-1 target gene

All experiments performed in cortical neurons. A) Deletion analysis of a luciferase-based reporter of the Sesn2 promoter. B) Effect of an interfering mutant of C/EBP on activity-dependent induction of the Sesn2 reporter construct. *p<0.05 compared to control plasmid (n=5). C) Effect of putative C/EBP binding site mutation on activity-dependent induction of the Sesn2 promoter (distal site ATTTCACACC mutated to ATTTCGGCCC; proximal site TTTGCAGCATC mutated to TTTGCGGCCTC). NB. C/EBP sites have consensus (A/T)TTGCG(C/T)AA(C/T), although quite substantial variations are tolerated⁵⁰. D) q-RT-PCR of C/EBPβ induction by synaptic activity at 4 h (n=6). E) Deletion analysis of a luciferase-based reporter of the Srxn1 promoter. F) Effect of AP-1 (cfos+cjun) expression on activity of the Srxn1 reporter construct, both wild-type and that containing mutations in the two putative AP-1 binding sites (distal site TGAGTCA mutated to TAAGCTT, proximal site TGAGTCA mutated to TGGGCCC). G) Effect of AP-1 site mutation on activity-dependent induction of the Srxn1 promoter.

Transcription of the Cebpb gene (but not Cebpa) is induced by BiC/4-AP stimulation (Fig. 6d, and data not shown), raising the possibility that this new expression may be important for Sesn2 induction. Consistent with this, Sesn2 is not an immediate early gene: it was not induced at 1 h (Fig. S5b), and induction at 4 h was severely inhibited by the protein synthesis inhibitor cycloheximide (Fig. S5b). C/EBPβ is a CREB-regulated immediate early gene²⁷, induction of which is not inhibited by cycloheximide (Fig. S5c). The rapid induction at 1 h contrasts with the slower induction of Sesn2, consistent with the predominantly delayed-response nature of Sesn2 induction.

Srxn1 is an AP-1 target gene

We created a Srxn1-Luciferase reporter which was activated by BiC/4-AP-induced NMDAR activity (Fig. S5a). Deletion analysis identified the activity-inducible region between -645 and -234 relative to the translation start site (Fig. 6e). Analysis of this region revealed two putative AP-1 sites (TGAGTCA). Ca²⁺ signaling is known to activate AP-1 family members, confirmed by our observation that BiC/4-AP stimulation activates a heterologous AP-1 reporter 6.1 ± 0.61 fold (n=4). Overexpression of AP-1 (c-fos and c-jun) strongly induced activity of the Srxn1 reporter, which was impaired when either putative AP-1 site was mutated, and abolished in the double mutant (Fig. 6f). Moreover, mutation of either site significantly reduced activity-dependent induction of the Srxn1 reporter construct, and the double mutation abolished induction (Fig. 6g). Thus, Srxn1 is activated via AP-1 sites in its promoter.

Extrasynaptic NMDARs fail to boost antioxidant defences

Trans-synaptic activation of synaptic NMDARs is anti-apoptotic, however, chronic activation of all (synaptic and extrasynaptic) NMDARs by bath application of glutamate/NMDA is not¹⁷. Moreover, synaptic vs. bath activation of NMDARs promote very different transcriptional responses²⁸. We tested whether trans-synaptic activation of synaptic NMDARs is critical for NMDAR-dependent antioxidative effects, or whether bath application of glutamate will suffice. We bath-applied a range of glutamate doses (up to the toxicity threshold: 20 μM) in the presence of TTX to ensure that there was no preferential activation of synaptic NMDARs, as can occur²⁹. Stimulations were compared to control (TTX only).

We first analysed bath-glutamate signaling to Akt, FOXO export, and Txnip suppression. Sustained glutamate exposure only triggered transient Akt activation (Fig. 7a) and weak FOXO export (Fig. 7b). Suppression of Txnip expression was poor compared to BiC/4-AP stimulation (Fig. 7c). Also, no dose of bath glutamate induced either Sesn2 or Srxn1 appreciably (Fig. 7d,e) nor was there significant induction of Sesn2’s activator, Cebpb (Fig. 7f). Consistent with this, glutamate did not promote the reduction of overoxidized Prxs (Fig. 7g,h), nor induce protection against H₂O₂-induced death (Fig. 7i). Thus, the strong antioxidative effect of NMDAR signaling is restricted to the trans-synaptic activation of synaptic NMDARs, and is not a pre-conditioning-type effect in response to sub-toxic glutamate exposure.

Open in a separate window

Fig. 7

Extrasynaptic NMDARs do not promote antioxidative effects

A) Phospho-(Serine 473)-Akt kinetics in cortical neurons in response to a range of glutamate concentration (normalized to Akt levels, n=3). The dashed line indicates the level of phospho-Akt induced by BiC/4-AP treatment at 2 h. Example Western also shown. B) Subcellular distribution of transfected myc-tagged FOXO1 analyzed by immunofluorescence, and stimulated as indicated for either 30 min or 2 h (n=3). C-F) q-RT-PCR of the indicated genes in response to a range of glutamate concentrations (4 h stimulation). Dashed line indicates the level of expression following 4 h BiC/4-AP stimulation (n=3). G,H) Lack of recovery of PrxII-SO_2/3H following a 1 h exposure to high [H₂O₂] (200 μM). Neurons treated with glutamate for 12 h prior to H₂O₂ exposure. Washout (w/o) period is 16 h. Prx-SO_2/3H levels normalized to Prx loading (n=3). I) Cell death due to 24 h H₂O₂ insult in the face of the indicated glutamate treatments, applied 12 h before insult (n=3).

We next investigated whether these differences are due to NR2 subunit differences between synaptic and extrasynaptic NMDARs, since NR2A-containing NMDARs may be enriched at synaptic locations³⁰ and signal to neuroprotection more effectively than NR2B-containing NMDARs³¹. Both whole-cell NMDAR currents and extrasynaptic NMDAR currents (analyzed by blocking synaptic NMDARs with the “quantal block” protocol, Fig. S1c) were blocked by 70% by the NR2B antagonist ifenprodil (Fig. 8a). Ifenprodil does not completely block responses even when mediated exclusively by NR2B-containing NMDARs (around 80 % is achieved³²). Thus, synaptic and extrasynaptic NMDARs are overwhelmingly and equally NR2B-dominated, a likely consequence of the relatively early developmental stage under study.

Open in a separate window

Fig. 8

Memantine, but not NR2B antagonists, discriminate between pro-survival and pro-death NMDAR signaling

A) Ifenprodil (3 μM) sensitivity of whole-cell and extrasynaptic currents (n=8) isolated using the “quantal block” protocol. B) Cell-death induced by 1 h exposure to 50 μM NMDA ± NMDAR antagonists MK-801 (10 μM), ifenprodil (3 μM), Ro 25-6981 (500 nM) *p<0.05 (n=3). C) Cell death due to 24 h H₂O₂ in the face of the indicated treatments, applied 12 h before insult. *p<0.05 (n=4). D) Txnip mRNA expression in neurons treated as indicated (24 h). *p<0.05 (n=3). E) Cell death due to 24 h H₂O₂ in the face of the indicated treatments. *p<0.05 (n=3). F) Effect of ifenprodil on BiC/4-AP induction of Sesn2-Luc and Srxn1-Luc. *p<0.05 (n=4). G) Effect of memantine (5 or 10 μM) on NMDA (50 μM)-induced cell death. *p<0.05 compared to control NMDA treatment (n=3). H) Cell death due to 24 h H₂O₂ insult in the face of the indicated treatments. *p<0.05 (n=3). I) Subcellular distribution of transfected myc-tagged FOXO1 analyzed by immunofluorescence. J) Txnip mRNA expression in neurons treated for 24 h as indicated. K-M) Effect of memantine on BiC/4-AP-induced changes in Sesn2 (K), Srxn1 (L) and Txnip (M) expression (n=3). N) Effect of memantine on BiC/4-AP-induced protection against H₂O₂ treatment. O) TUNEL analysis of the cortex of P7 mice subjected to IP-injection of memantine (20 mgkg^-1) at P6 (for 24 h, n=6). Dotted line indicated the level of death observed with MK-801 injection (Fig. 1a).

We hypothesized that both harmful and protective consequences of NMDAR activation in these neurons would be inhibited by NR2B antagonists. Both ifenprodil and Ro 25-6981 protected neurons from cell death induced by a toxic dose of NMDA (50 μM, Fig. 8b). NR2B antagonists also impaired protective signaling by trans-synaptic activation of synaptic NMDARs: ifenprodil and Ro 25-6981 exacerbated neuronal death by H₂O₂ (Fig. 8c), caused an up-regulation of Txnip (Fig. 8d), reduced BiC/4-AP induced protection against H₂O₂-induced death (Fig. 8e) and reduced signaling to Sesn2 and Srxn1 activation (Fig. 8f). Thus, differences in the effect of synaptic stimulation vs. bath application of agonist are unlikely to be due to large, clear-cut differences in NR2 subunit composition at synaptic and extrasynaptic sites. However, we cannot rule out a role for NR2A-containing NMDARs at the synapse-this would require analyses of NR2A-null neurons.

Memantine is an open channel blocker with a fast off-rate. Its uncompetitive nature results in an effective blockade of chronic NMDAR activity caused by bath-applied glutamate³³. However, due to its fast off-rate and voltage-dependent binding properties, memantine does not substantially interfere with normal synaptic NMDAR activity by accumulating in the channel³³. In the context of this study, memantine appears suited to the sparing of the antioxidative effects of trans-synaptic activation of synaptic NMDARs, while blocking chronic activation of NMDARs by low level glutamate which does not enhance antioxidant defences (Fig. 7i).

Memantine (at 5 μM or 10 μM) blocks NMDAR-dependent neuronal death induced by 50 μM NMDA (Fig. 8g). However, it does not interfere with trans-synaptic NMDAR-dependent activation of antioxidant defenses: unlike MK-801, memantine does not exacerbate H₂O₂-induced death (Fig. 8h) nor promote nuclear localisation of FOXOs (Fig. 8i) or Txnip expression (Fig. 8j), and does not interfere with synaptic activity-dependent signaling to Sesn2 and Srxn1 activation (Fig. 8k,l) nor Txnip suppression (Fig. 8m), nor activity-dependent neuroprotection against H₂O₂ (Fig. 8n). We also assessed the effects of memantine administration in vivo at a dose (20 mgkg^-1) consistently shown to be therapeutically active in protecting against ischemic insults ³⁴ and known to result in brain concentrations of memantine similar to those used in our in vitro experiments³⁵. Unlike MK-801, memantine injection does not cause neuronal apoptosis in the P7 cortex: no death was observed either by TUNEL-staining (Fig. 8o) or Fluorojade staining, which labels degenerating neurons irrespective of mechanism of death (data not shown). The differential effects of MK-801 and memantine emphasize that trans-synaptic stimulation of synaptic NMDARs is important for boosting antioxidant defences and that blockade of synaptic NMDARs boosts vulnerability to oxidative stress in vitro and in vivo.

Txnip promotes oxidative stress and cell death

Txnip interacts with the reduced form of thioredoxin, inhibits its biochemical activity and sensitizes a variety of cell types to H₂O₂-induced death^{3, 20}. Txnip is implicated in the pathogenesis of diabetes: it is up-regulated in the vasculature and pancreatic beta cells of diabetic rats and is induced by hyperglycaemia in both systems, promoting oxidative stress^{20, 36}. Txnip plays a role in oncology: it is proposed to suppress tumour cell growth and metastasis, and reduced levels have been observed in a variety of cancers³. Txnip’s influence on neuronal vulnerability to oxidative stress is consistent with its role as a pro-oxidative gene in non-neuronal systems.

Studies on transcriptional control by synaptic Ca²⁺ signals focus almost exclusively on up-regulated genes. Results here concerning the regulation of Txnip expression demonstrate that gene inactivation must also be considered. Our findings that Txnip is regulated by FOXOs adds to our knowledge of FOXO-regulated pro-death genes, which thus far have focussed mainly on Bim and Fasl​^{24, 37}.

Peroxiredoxins are cytoprotective antioxidant enzymes

Prxs are a family of thiol-based antioxidants with cytoprotective effects in the face of oxidative insults. PrxII protects cortical neurons against Aβ toxicity³⁸ and oxygen-glucose deprivation ³⁹. Interfering with PrxII expression renders neuroblastoma cells vulnerable to oxidative stress⁵, and renders cortical neurons vulnerable to MPP⁺​ ⁶. PrxIII protects hippocampal neurons against excitotoxicity⁴.

Low reported catalytic rates for Prxs led to the suggestion that Prxs would not be able to compete with catalase or glutathione peroxidases for cellular peroxide, despite evidence of the antioxidative properties of Prxs. This apparent paradox was due to an underestimation of peroxidase rates of mammalian PrxII, as a result of their indirect calculation by measuring NADPH oxidation rates which measure regeneration by the thioredoxin system. When peroxide levels are high, it is clear that the thioredoxin system cannot keep up. In fact, when rates of mammalian Prx were measured directly⁴⁰ they were found to be 100 fold greater than previously thought, similar to rates recently found in PrxI and II from S. cerevisiae​⁴¹. Since Prxs are relatively abundant, they are likely to play an important antioxidant role in mammalian in vivo systems.

Thus, molecular events that render Prxs inactive in vivo in pathological scenarios may contribute to disease progression. We observe Prx-inactivating Prx-SO_2/3H formation following cerebral ischemia in vivo (Fig. 9), but this is not the only way by which Prxs can be inhibited: The enzymic activity of PrxII can be impaired by cdk5-mediated phosphorylation⁶ and by nitrosylation⁴², both of which occur in Parkinson’s Disease^{6, 42}.

Re-activation of Prx-SO_2/3H by Sulfiredoxin and Sestrin2

Sulfiredoxin was initially characterized in yeast⁴³ and then in mammalian cells ⁸ and acts by catalysing the ATP-dependent formation of a sulfinic acid phosphoric ester on Prx⁸ which is then reduced by thiol equivalents such as thioredoxin. The mechanism of ATP-dependent Sesn2 is unknown⁹. The capacity of neurons to reduce inactive Prx-SO_2/3H was hitherto unknown, and our findings show that active neurons have a far greater capacity for this than inactive neurons, and are associated with high levels of Srxn1 and Sesn2. This is the first time that Prx-SO_2/3H reducing activity has been found to be signal-dependent. Given the widespread expression of these genes, this raises the question as to whether the Prx-SO₂H reduction process is signal-inducible in vivo. For example, ischemic preconditioning is associated with activation of AP-1⁴⁴ which could therefore result in up-regulation of Srxn1 expression and reduce Prx-SO_2/3H formation that occurs following ischemia (Fig. 9).

Neuronal cultures, stimulation, and the induction of oxidative stress

Cortical rat neurons were cultured as described ¹⁸ from E21 rats or E17 mouse pups in Neurobasal-A medium and B27 (Invitrogen). Stimulations and transfections (Lipofectamine 2000, Invitrogen) were done at DIV8-10 after transferring neurons into trophically-deprived medium ¹⁸. Action potential bursting was induced by treatment with 50 μM bicuculline, plus 250 μM 4-aminopyridine (Sigma) to enhance burst frequency. Stimulations were initiated 12 h prior to the application of an oxidative insult. Neurons were fixed after a further 24 h and subjected to DAPI staining and cell death quantified by counting (blind) the number of apoptotic nuclei as a percentage of the total. For details of analysis of peroxide-induced ROS accumulation and caspase activity, see supplemental methods.

Electrophysiological recording and analysis

Currents were recorded (at 21±2°C) using an Axopatch-1C amplifier (Molecular Device, Union City, CA). Data were digitized and analyzed using WinEDR v6.1 software (John Dempster, University of Strathclyde, UK). Neurons were voltage-clamped at -70 mV, and recordings were rejected if the holding current was >-100 pA or if the series resistance drifted by >20% of its initial value (<35 MΩ). See supplemental methods for details of solutions, drugs and of “activity block” and “quantal block” protocols

Gene expression analysis by qPCR and microarray

For qPCR, cDNA was synthesized from 1-3 μg RNA (extracted using Qiagen Rneasy kit) using the Stratascript QPCR cDNA Synthesis kit and qPCR performed in an Mx3000P QPCR System (Stratagene) using Brilliant SYBR Green QPCR Master Mix. For detailed protocols and primer sequences, see supplemental methods. For microarray-based expression analysis, total RNA was quality-confirmed on RNA 6000 Nanochips in the Agilent 2100 Bioanalyzer. Purified biotinylated target cRNA was produced (see supplemental methods) and hybridized to Mouse Genome 430A plus 2.0 microarrays (Affymetrix). After hybridization, the arrays were washed and scanned (Genechip Scanner 3000, Affymetrix). Expression was calculated using the robust multiarray average algorithm implemented in the Bioconductor (http://www.bioconductor.org) extensions to the R statistical programming environment.

In vivo procedures

In vivo MK-801 and memantine administration, TUNEL histology, and protein carbonyl assay of extracted proteeins were all performed using established techniques. In vivo focal cerebral ischemia was performed on adult male C57Bl/6J mice under an appropriate Home Office Licence. See supplemental methods section for details of all in vivo procedures.

We thank Peter Brophy for critically reading the manuscript, and acknowledge Janine Stuwe’s assistance. We also thank David Bennett and the Rush Alzheimer’s Disease Center (NIH grant P30AG10161) for providing some of the brain samples used in this study, and thank Domenico Accili, Hilmar Bading, Jean-Claude Chambard, Richard Lee, Charles Vinson, George Wilding and Junji Yodoi for plasmids. This work was funded by the Wellcome Trust, a Royal Society University Research Fellowship (GH), Medical Research Scotland, Tenovus Scotland, the BBSRC, Sanitaetsrat Dr. Emil Alexander Huebner and Gemahlin-Stiftung and a Rahel Hirsch scholarship from the Humboldt University Berlin.