TRAF1 transgenic mice exhibit enlarged stroke lesions

To elucidate the direct effects of upregulated neuronal TRAF1 expression on stroke outcome, we generated several lines of neuron-specific Traf1 transgenic (TG-TRAF1) mice. TRAF1 expression in neurons is driven by the platelet-derived growth factor promoter (Supplementary Fig. S2a), and the genotypes of the TG-TRAF1 mice were confirmed through PCR (Supplementary Fig. S2b). To reflect physiological conditions during stroke, TG2-TRAF1 mice, which exhibit levels of cerebral TRAF1 expression similar to those induced by MCAO in vivo (Supplementary Fig. S2c), were used for further experiments. In addition, all mouse strains were normal in appearance and fertile. To evaluate whether TRAF1 neuron-specific overexpression resulted in phenotypic changes to the cerebral vasculature, we inspected the Circle of Willis, anterior cerebral arteries, middle cerebral arteries (MCAs) and posterior cerebral arteries of the TG2-TRAF1 and non-transgenic (NTG) mice using India ink staining. We did not observe gross anatomical differences that might affect stroke outcome (Supplementary Fig. S2d), so to eliminate the possibility that TRAF1 might have altered cerebral perfusion and impacted infarct volumes, regional cerebral blood flow (rCBF) was monitored by Doppler flowmetry throughout the MCAO procedure. TG2-TRAF1 mice exhibited no significant difference in rCBF compared with NTG mice at baseline (Supplementary Fig. S2e). Ligation of the common carotid artery resulted in a comparable reduction in blood flow in both experimental groups, and rCBF in the MCA territory returned to baseline levels after reperfusion (Supplementary Fig. S2e). Because both hypertension and arrhythmia are particularly notable stroke risk factors2, we then measured blood pressure and heart rate in both strains; however, no differences in systolic blood pressure, diastolic blood pressure or heart rate were observed between the groups (Supplementary Table S1); and furthermore, no significant changes were noted in blood gases, pH or body weight.

TG2-TRAF1, TG4-TRAF1 and NTG mice were challenged with tMCAO for 60min and examined for 2,3,5-triphenyl-2H-tetrazolium chloride (TTC) staining after 23 or 71h of reperfusion (Fig. 2a,b). Compared with their NTG siblings, a 6.7-fold increase in TRAF1 expression in TG2-TRAF1 mice enlarged the infarct volume by ~37% and 53% (Fig. 2c) and the oedema volume by ~47% and 34% (Fig. 2d) at 24 and 72h, respectively, after MCAO. TG2-TRAF1 mice also had significantly higher neurological scores at both time points, with increases of ~26% and 27% at 24 and 72h, respectively (Fig. 2e). TG4-TRAF1 mice displayed an ~9.2-fold increase in TRAF1 expression compared with NTG mice and displayed outcomes similar to TG2-TRAF1 mice (Fig. 2c–e). Notably, no significant change in cerebral TRAF1 expressions was observed in both TG2 and TG4 mice before and after I/R insults (Supplementary Fig. S2f). Neurons have limited ability to regenerate and are particularly at risk for programmed cell death due to their high metabolic demand during ischaemic brain injury3. Thus, we evaluated neuroapoptosis by performing TUNEL assays and NeuN dual-immunofluorescence staining 24h after transient focal ischaemia and found an increase of ~41% in the apoptotic rate in TG2-TRAF1 mice compared with that in NTG mice (Fig. 2f). Thus, these data demonstrate that neuronal TRAF1 has a detrimental effect on both histological cerebral ischaemic injury and behavioural neurological dysfunction, which is most likely attributable to the enhancement of delayed neuronal apoptosis.

Open in a separate window

Figure 2

TRAF1 neuron-specific transgenic mice have enlarged brain lesions after stroke.

(a) Time course of MCAO and reperfusion before sacrifice. (b) Brains of NTG, TG2-TRAF1 and TG4-TRAF1 mice stained with TTC at the indicated times after I/R. Scale bar, 10mm. (c–e) Quantification of infarct (c), oedema (d) volumes and neurological scores (e) at the indicated times (n=7 at each time point, *P<0.05 versus NTG). (f) Representative images of apoptosis in cortical neurons co-stained with NeuN and TUNEL after 24h of reperfusion (left). Scale bar, 100μm. Insets show a higher magnification view. Scale bar, 20μm. The right panel shows the quantification of TUNEL-positive cells (n=4, *P<0.05 versus NTG). The data represent the mean±s.d. Statistical analysis was carried out by unpaired Student’s t-test.

TRAF1-deficient mice have reduced stroke lesions

We hypothesized that if TRAF1 has a detrimental role during stroke, then inhibiting the post-stroke generation of TRAF1 in Traf1 knockout mice (TRAF1-KO) might prevent neuronal damage. All mouse strains were generated in agreement with predicted Mendelian ratios (85 wild-type (WT), 152 heterozygous and 81 TRAF1-KO mice). As expected, the TRAF1 protein was not expressed in the brains of TRAF1-KO mice before or after MCAO (Supplementary Fig. S3a). Likewise, we ruled out potential changes in gross neurovascular anatomy (Supplementary Fig. S3b), rCBF (Supplementary Fig. S3c) and other physiological parameters, including blood pressure, blood gases, pH and body weight (Supplementary Table S1). Importantly, TRAF1 deficiency led to ~38% and 39% reductions in infarct and oedema volumes, respectively, compared with those in WT mice 24h after MCAO (Fig. 3a–c). These neuroprotective effects on the infarct and oedema volumes persisted until 71h after reperfusion (35% and 49% reductions, respectively, compared with WT; Fig. 3b,c). In addition, the neurological score was mitigated by >30% in TRAF1-KO mice at both time points (Fig. 3d). Consistent with these observations, TRAF1-KO mice displayed an ~57% reduction in TUNEL-positive cells 24h after I/R (Fig. 3e), further indicating a critical role of TRAF1 in ischaemic brain injury. Therefore, inhibiting TRAF1 induction after cerebral ischaemia could be neuroprotective.

Open in a separate window

Figure 3

TRAF1 deficiency is neuroprotective during stroke.

(a) Brains stained with TTC at the indicated times after ischaemia/reperfusion (I/R). Scale bar, 10mm. (b–d) Quantification of infarct (b) and oedema (c) volumes and neurological scores (d) at the indicated times (n=7 or 12 at each time point, *P<0.05 versus WT). (e) Representative images of apoptosis in cortical neurons of WT and TRAF1-KO mice co-stained with NeuN and TUNEL after 24h of I/R (left). Scale bar, 100μm. The inset shows a higher magnification view. Scale bar, 20μm. The right panel shows quantification of TUNEL-positive cells (n=4, *P<0.05 versus NTG). The data represent the mean±s.d. Statistical analysis was carried out by unpaired Student’s t-test.

TRAF1 promotes ischaemia-induced neuronal apoptosis

Given that neuronal apoptosis has been suggested to underlie progressive neuropathology and neurological dysfunction hours or days post-stroke18, we investigated the biological role of TRAF1 in neuroapoptotic cascades in vivo. We first examined caspase-3, which is a pivotal effector of apoptosis in ischaemic stroke4, and found that the number of cleaved (active) caspase-3-positive cells was ~31% lower in TRAF1-KO mice and ~2.1-fold greater in TG2-TRAF1 mice than in controls (Fig. 4a). TRAF1 deficiency also upregulated the mRNA and protein expression levels of Bcl2 (a neuroprotective, anti-apoptotic gene4), whereas TRAF1 deficiency suppressed the expression of the pro-apoptotic genes Bax, Fas and FasL (the last two are crucial mediators of the extrinsic mechanism4) 6h after MCAO (Fig. 4b–d). Thus, these results suggest that TRAF1 is an upstream modulator of the mitochondrial apoptotic pathway in neurons. To further validate that TRAF1 accelerates programmed neuronal death, we investigated whether TRAF1 directly affects neuronal survival in cultured cortical neurons. We performed gain- and loss-of-function experiments using an adenovirus harbouring human TRAF1 cDNA (Ad-TRAF1) or TRAF1 short hairpin RNA (Ad-shTRAF1). Under basal conditions, no significant difference in neuronal survival was observed between the treatments. However, when challenged with OGD, Ad-TRAF1 infection, which increased TRAF1 expression by ~2.4-fold (Fig. 4e), led to fewer viable neurons (Fig. 4f) and increased lactate dehydrogenase (LDH) release (Fig. 4g) at the different time points. By contrast, Ad-shTRAF1 treatment, which decreased TRAF1 expression by ~88% (Fig. 4e), significantly counteracted OGD-induced cell death (Fig. 4f,g). Consistent with these results, the positive and negative regulation of pro- and anti-apoptotic proteins, respectively, by TRAF1 in OGD-challenged cells (Supplementary Fig. S4a,b) was similar to that observed in vivo. Therefore, our results show that TRAF1-mediated ischaemic cerebral injury is, at least in part, attributable to delayed neuronal cell death. Thus, targeting TRAF1 may render neurons resistant to programmed cell death following ischaemia–reperfusion.

Open in a separate window

Figure 4

TRAF1 modulates neuronal apoptosis in ischaemic brain.

(a) Fluorescence staining of cleaved caspase-3 (red) and nuclei (DAPI, blue) in the brains of WT, TRAF1-KO, NTG, and TG2-TRAF1 mice after 24h of I/R. Scale bar, 100μm. The inset shows a higher magnification view. Scale bar, 20μm. The right panel shows the quantification of immunopositive cells (n=3–4, *P<0.05 versus WT, ^#P<0.05 versus NTG). (b) Quantitation of the indicated mRNA levels (n=4, *P<0.05 versus WT, ^#P<0.05 versus NTG). (c,d) Immunoanalysis of pro-apoptotic proteins in the brain extracts of WT and TRAF1-KO (c) and TG2-TRAF1 and NTG (d) mice 6h after sham surgery or I/R. The right panels show the quantification of normalized protein levels (n=6, *P<0.05 versus sham-operated WT (c) or NTG (d); ^#P<0.05 versus I/R-operated WT (c) or NTG (d)). (e) Immunoanalysis of TRAF1 in neurons infected with the indicated adenoviral vectors (left) and quantification of TRAF1 expression normalized to GAPDH (right) (n=6, *P<0.05 versus Ad-shRNA, ^#P<0.05 versus Ad-GFP). (f,g) Cell viability (f) and LDH release (g) after adenoviral infection of neurons challenged by OGD for the indicated times (n=6 for each time point, *P<0.05 versus Ad-GFP, ^#P<0.05 versus Ad-shRNA). The data represent the mean±s.d. Statistical analysis was carried out by unpaired Student’s t-test.

TRAF1 promotes ischaemic brain injury via ASK1 activation

On the basis of the observation that TRAF1 activates the Fas/FasL apoptotic pathway, we proposed that TRAF1 functions upstream of the mitochondria-mediated apoptosis signalling pathway. The MAPK family, which consists of p38 MAPK, ERK and JNK, is increasingly being recognized for its involvement in controlling post-ischaemic neuronal cell death1. Therefore, we investigated whether these pathways are modulated by TRAF1 upon ischaemic insults. TRAF1 exerted no significant effect on either the expression or phosphorylation of ERK and p38 (Supplementary Fig. S5a,b). JNK, which has been shown to be markedly upregulated 6h after cerebral ischaemic injury in an experimental stroke model19, is known to mediate the induction of a series of pro-apoptotic proteins, including c-Jun, Bim, Bax and Fas, upon I/R20. In our model, TRAF1-KO mice displayed reduced phosphorylation of JNK and its major downstream molecule, c-Jun, at serines 63 and 73 (Fig. 5a). Because JNK is activated by a number of upstream stress-related activators, including ASK1 and MKK4/7 (ref. 21), we examined whether ASK1 activation was altered; we observed an ~36% decrease in the expression of active, phosphorylated ASK1 (p-ASK1), concurrent with inhibition of the MKK/JNK pathway (Fig. 5a). Conversely, we found that TRAF1 overexpression increased p-ASK1 expression to ~1.76-fold greater than that in NTG controls and enhanced the activation of the MKK/JNK pathway (Fig. 5b), thus aggravating I/R-induced cerebral injury. To verify the positive regulation of the ASK1/JNK pathway by TRAF1, we used gain- and loss-of-function approaches. Western blot analysis confirmed that ASK1/JNK activation was positively regulated by TRAF1 upon OGD stress in vitro (Supplementary Fig. S6). Notably, inhibition of TRAF2 did not interfere with TRAF1-mediated ASK1 phosphorylation during OGD/reperfusion (Supplementary Fig. S7). Thus, TRAF1 may underlie, at least in part, post-stroke neuroapoptosis through the ASK1/JNK pathway. Another crucial signalling pathway involved in I/R upstream of the mitochondrial cascade is the Akt pathway. Akt promotes cell survival via phosphorylation and activation of the pro-survival mTOR and CREB pathways22. Furthermore, Subramanyam et al.23 observed that ASK1 inhibits Akt phosphorylation and activation and participates in the apoptotic inhibition of the regulatory volume increase in HeLa cells. However, whether Akt protects against neuronal death downstream of ASK1 is not known. We found that TRAF1-induced ASK1 activation coincided with reduced phosphorylation of Akt and the downstream molecule mTOR both in vivo (Fig. 5c,d) and in vitro (Supplementary Fig. S8a,b). The activity of the Akt/mTOR signalling pathway was significantly enhanced in the brains of TRAF1-KO mice after MCAO as well as in neurons infected with Ad-shTRAF1 and exposed to OGD. Akt phosphorylates CREB, leading to CREB-mediated expression of genes crucial for neuronal survival, including brain-derived neurotrophic factor (BDNF). The CREB/BDNF pathway was activated in TRAF1-KO mice following MCAO, resulting in increased expression of the pro-survival factors GAP43 and NGF (Supplementary Fig. S9a). Conversely, TG2-TRAF1 suppressed CREB and BDNF phosphorylation and decreased GAP43 levels (Supplementary Fig. S9b). Together, these findings indicate that TRAF1 confers pro-apoptotic functions via regulation of both the JNK and Akt pathways and that this regulation is most likely mediated by ASK1 activation.

Open in a separate window

Figure 5

TRAF1 activates ASK1 signalling pathways.

(a,b) TRAF1 activates the ASK1/JNK pathway. Western analysis (left) and quantification (right) of the levels of the indicated proteins in the brains of (a) TRAF1-KO, WT and (b) TG2-TRAF1, NTG mice 6h after sham surgery or ischaemia/reperfusion (I/R) (n=6, *P<0.05 and ^#P<0.05 versus WT and NTG with and without I/R, respectively). (c,d) TRAF1 suppresses the Akt cell survival pathway. Analysis of the indicated proteins in mice under the same experimental conditions described in a and b, respectively (n=6). The data represent the mean±s.d. Statistical analysis was carried out by unpaired Student’s t-test.

TRAF1 directly interacts with ASK1

Next, we examined whether ASK1 is a direct target of TRAF1. We found that in HEK293T cells transfected with Myc-tagged TRAF1 and Flag-tagged ASK1, TRAF1 co-immunoprecipitated with ASK1, and vice versa (Fig. 6a). The results of our glutathione S-transferase (GST) pull-down assay (Fig. 6b) and the endogenous co-immunoprecipitation (co-IP) performed in primary cortical neurons (Supplementary Fig. S10) further supported the finding that TRAF1 interacts with ASK1. Next, we investigated TRAF1 and ASK1 localization and observed that these proteins were predominantly co-localized in the cytoplasm (Fig. 6c). To identify the protein domains critical for TRAF1 and ASK1 association, we generated a series of truncated Flag-tagged ASK1 constructs (Fig. 6d) and co-transfected ASK1 deletion mutants with Myc-tagged TRAF1 followed by co-IP. The results reproducibly demonstrated that either the N-terminal region or kinase domain of ASK1 but not the C-terminal region retained the ability to interact with TRAF1 (Fig. 6e). TRAF1 comprises a C-terminal (TRAF) domain responsible for interactions with receptors and other signalling proteins24 and an N-terminal domain that lacks the RING finger domain present in other TRAFs. Using TRAF-domain and N-terminal-domain truncation and deletion mutants (Fig. 6d), we found that only the TRAF domain bound to ASK1 (Fig. 6f). To determine whether this binding domain directly mediates TRAF1-induced ischaemic neuronal injury, we generated an adenoviral vector encoding TRAF1 with a mutant TRAF domain (Ad-mutant). Indeed, 24h after OGD treatment, cultured neurons infected with Ad-TRAF1, but not those infected with Ad-mutant, displayed a reduction in cell viability of ~37% (Fig. 6g) and an increase in LDH release by 32% (Fig. 6h) compared with the Ad-GFP control. Therefore, the TRAF domain is required for TRAF1-mediated neuronal death following ischaemia. Notably, simple genetic TRAF1 manipulations failed to impact ASK1 activation in sham-operated mice (Fig. 5a,b) and control neuronal cultures (Supplementary Fig. S6a,b). We thus speculated that I/R exposure is necessary for TRAF1-ASK1 interaction and, subsequently, ASK1 phosphorylation. To test this hypothesis, we isolated primary cortical neurons and infected them with Ad-TRAF1 prior to OGD exposure. The protein extract was immunoprecipitated with anti-TRAF1 antibodies and immunoblotted with anti-ASK1 antibodies, and vice versa (Supplementary Fig. S11). Intriguingly, although the TRAF1-ASK1 interaction was relatively weak in normal neuronal cultures, TRAF1 readily bound to ASK1 after an OGD challenge (Supplementary Fig. S11). Therefore, our results indicate that in response to I/R, TRAF1 expression is induced in neurons, and the ability of TRAF1 to interact with ASK1 is improved.

Open in a separate window

Figure 6

TRAF1 directly interacts with ASK1.

(a) Co-IP of TRAF1 and ASK1. HEK293T cells were transfected with Myc-tagged TRAF1 and FLAG-tagged ASK1. The lysates were immunoprecipitated with anti-Myc or anti-FLAG and analysed by immunoblotting using anti-Myc or anti-FLAG antibodies. (b) Immunoanalysis of GST pull-down of FLAG-tagged ASK1 with GST-TRAF1 or GST. (c) Co-localization of TRAF1 and ASK1 in the cytoplasm of HEK293T cells; the nuclei were stained with DAPI. Scale bar, 5μm. (d) Schematic representation of ASK1 and TRAF1 deletion mutants. (e) Mapping of the TRAF1-binding region of ASK1. HEK293T cells were transfected with Myc-tagged TRAF1, and FLAG-tagged ASK1 deletion mutants were immunoprecipitated from lysates with anti-FLAG and immunoblotted with anti-Flag and anti-Myc. (f) Mapping of the ASK1 binding region of TRAF1. HEK293T cells were transfected with FLAG-tagged ASK1, and the lysates were immunoprecipitated with anti-Myc and immunoblotted with anti-Flag and anti-Myc. (g,h) TRAF1-ASK1 interaction is required for TRAF1-mediated neuronal injury. Cultured neurons were infected with Ad-GFP, Ad-TRAF1 or Ad-mutant harbouring a mutant TRAF domain and subject to OGD for the indicated times. The Ad-mutant completely rescued the Ad-TRAF1-mediated decrease in cell survival (g) and increase in LDH release (h) (n=9, *P<0.05 versus Ad-GFP; ^#P<0.05 versus Ad-TRAF1). The data represent the mean±s.d. Statistical analysis was carried out by unpaired Student’s t-test.

Mouse cerebral I/R model

Focal cerebral ischaemia was induced by occluding the left MCA using the intraluminal filament technique42,43,44. Briefly, the mice were anesthetized with 2.5–3% isoflurane in O₂. The rectal temperature was maintained at 37±0.5°C using a homoeothermic blanket. rCBF was measured using a laser-Doppler flowmetry instrument (Periflux System 5010; Perimed, Järfälla, Sweden) with a flexible probe affixed to the skull (2mm posterior and 5mm lateral to the bregma). To achieve tMCAO, a 6-0 silicon-coated monofilament surgical suture (Doccol, Redland, CA, USA) was inserted into the left external carotid artery, advanced into the internal carotid artery and wedged into the cerebral arterial circle to obstruct the origin of the MCA. A decline in rCBF >75% confirmed the interruption of blood flow in the MCA. Sixty minutes after MCAO induction, the filament was withdrawn, and a return to >70% of basal cerebral blood flow within 10min of suture withdrawal confirmed reperfusion of the MCA territory. The mice underwent reperfusion for the durations indicated in the text. The mice in which the filament was withdrawn immediately after the decline in rCBF were used as sham controls. Heart rate, systolic blood pressure and diastolic blood pressure were monitored in randomly selected, conscious mice using the non-invasive tail-cuff method. We measured body weight and arterial blood gases before surgery.

The experimental groups were WT (n=74), TRAF1 KO (n=70), NTG (n=56) and TG-TRAF1 (n=62). Analyses were conducted at the times after reperfusion indicated in the text. In all experiments, the examiners were blinded to the mouse genotypes.

India ink staining

The integrity of the cerebral vasculature of the experimental mice was confirmed by India ink staining42,43,44. Briefly, the animals were anesthetized by intraperitoneal injection of 50mgkg⁻¹ sodium pentobarbital (Sigma, St Louis, MO, USA) in 10ml physiological saline and perfused through the left cardiac ventricle with 2ml preheated staining solution containing 10% (W/V) gelatin (Amresco, Solon, OH, USA) and 50% (V/V) India ink (Solarbio, Beijing, China). Perfusion was stopped when the tissues, including the tongue, lips and gums, turned black. The isolated brains were immersed in 10% buffered formalin for 24h before examination. Vessels of the circle of Willis and their branches were photographed using a Nikon D700 digital camera.

Neurological deficit scores

Neurological deficits were assessed after 24 and 72h of reperfusion using a nine-point scale as follows: absence of neurological deficit, 0; left forelimb flexion upon suspension by the tail or failure to fully extend the right forepaw, 1; left shoulder adduction upon suspension by the tail, 2; reduced resistance to a lateral push towards the left, 3; spontaneous movement in all directions with circling to the left only if pulled by the tail, 4; circling or walking spontaneously only to the left, 5; walking only when stimulated, 6; no response to stimulation, 7; and stroke-related death, 8 (ref. 45).

Measurement of infarct volume

Following reperfusion, the mice were euthanized, and their brains were immediately removed. Mouse brains were cut into 1-mm coronal sections and a total of seven sections were prepared46. The sections were incubated with TTC staining solution for 15min at 37°C before fixed in 10% formalin solution overnight. The sections were imaged, and the volume of the infarct area was quantified with Image-Pro Plus 6.0. To exclude the effects of brain oedema, the percentage of infarct volume was calculated as follows: (volume of the contralateral hemisphere minus the volume of the non-lesioned ipsilateral hemisphere)/(contralateral volume times 2). We also calculated the relative oedema volume (%) as follows: (volume of the contralateral hemisphere minus the volume of the ipsilateral hemisphere)/contralateral volume. Finally, we calculated the infarct and oedema volumes by integrating their areas over seven brain sections.

Immunofluorescence and TUNEL staining

The animals were anesthetized with sodium pentobarbital and perfused with 0.1moll⁻¹ sodium phosphate buffer (pH 7.4) under 100mmHg pressure for 5min and with 4% paraformaldehyde in phosphate buffer for 15min. Immunocytochemistry was performed as previously described42. Briefly, the brains were rapidly removed and fixed in the same fixative solution for 6–8h at room temperature, followed by immersion overnight at 4°C in phosphate buffer containing 30% sucrose. The brains were embedded in OCT solution, and cryosections were prepared. The neurons were cultured on cover slips and fixed in ice-cold acetone. Next, the sections on cover slips were washed in PBS containing 10% goat serum and incubated overnight with the following primary antibodies at 4°C: mouse anti-NeuN (LV1825845; 1:200; Millipore, Temecula, CA, USA), chicken anti-MAP2 (Ab5392; 1:100; Abcam, Cambridge, UK), rabbit anti-cleaved caspase-3 (no. 19662; 1:100; Cell Signaling Technology, Danvers, MA, USA) and anti-TRAF1 (no. ab129279, 1:100, Abcam). The sections were washed in PBS and incubated in the appropriate secondary antibody for 1h. The secondary antibodies used were goat anti-chicken IgY (H&L, DyLight 488, ab96947, Abcam), goat anti-mouse IgG Alexa Fluor 568 conjugate (A11004; Invitrogen, Carlsbad, CA, USA) and anti-rabbit IgG Alexa Fluor 568 conjugate (A11011, Invitrogen). The nuclei were stained with 4',6-diamidino-2-phenylindole (DAPI). Images were acquired with a fluorescence microscope (OLYMPUS DX51) and DP2-BSW software (version 2.2), and the images were analysed with Image Pro Plus 6.0. For NeuN and TUNEL co-staining, the sections were first stained with NeuN antibody, followed by TUNEL staining with an ApopTag Plus In Situ Apoptosis Fluorescein Detection Kit (S7111, Millipore), according to the manufacturer’s protocol.

Tissue preparation

For quantitative real-time PCR and western blot analysis, the animals were anesthetized with 3% sodium pentobarbital and perfused via the left ventricle with cold sodium phosphate; the brains were then immediately removed. To collect the tissue in an unbiased manner that reflected the global extent of the infarcts, the olfactory bulbs and 1-mm sections of the anterior and posterior brain tissue were excised. We then collected the remaining tissues of the ipsilateral (including the infarct and peri-infarct areas) and contralateral (normal) hemispheres. All tissues were immediately frozen in liquid nitrogen and stored at −80°C.

Quantitative real-time PCR

Total RNA was isolated from frozen tissue using TRIZOL reagent (Invitrogen) following the manufacturer’s protocol, and 2μg of RNA was reverse-transcribed with the Transcriptor First Strand cDNA Synthesis Kit (Roche, Indianapolis, IN, USA). Quantitative RT-PCR analysis was performed using the LightCycler 480 SYBR Green 1 Master Mix (Roche) and the LightCycler 480 QPCR System (Roche). The PCR conditions were 95°C for 10min; 40 cycles of 95°C for 10s, 60°C for 10s and 72°C for 20s; and a final extension at 72°C for 10min. GAPDH served as the internal control. The following primer pairs were used:

Bax forward, 5′-TGAGCGAGTGTCTCCGGCGAAT-3′;

Bax reverse, 5′-GCACTTTAGTGCACAGGGCCTTG-3′;

Bcl2 forward, 5′-TGGTGGACAACATCGCCCTGTG-3′;

Bcl2 reverse, 5′-GGTCGCATGCTGGGGCCATATA-3′;

Fas forward, 5′-AATGGGGGTACACCAACCTGCG-3′;

Fas reverse, 5′-TTCACACGAGGCGCAGCGAA-3′;

FasL forward, 5′-GGGTGCCATGCAGCAGCCCATGAAT-3′;

FasL reverse, 5′-TGGCGGCAGTGGGAGTGGTTGTGAT-3′;

GAPDH forward: 5′-ACTCCACTCACGGCAAATTC-3′; and

GAPDH reverse: 5′-TCTCCATGGTGGTGAAGACA-3′.

Western blot analysis

Proteins were extracted from brains or cultured cells and homogenized in lysis buffer. Proteins (50μg) were separated on 8–12% SDS–PAGE gels and transferred to PVDF (polyvinylidene difluoride) membranes (Millipore, Bedford, MA). The membranes were blocked in TBST (Tris-buffered saline Tween-20) containing 5% skim milk powder for 1h at room temperature and incubated with primary antibodies overnight at 4°C. Next, the membranes were incubated with secondary antibodies (peroxidase-affinipure goat anti-mouse IgG (H+L) (115-035-003), or peroxidase-affinipure goat anti-rabbit IgG (H+L) (111-035-003); Jackson ImmunoResearch Laboratories, Lincoln, NE, USA) for 1h at room temperature. Membranes were treated with ECL reagents (170–5061; Bio-Rad) before visualization using FlourChem E imager (Proteinsimple, FlourChem E) according to the manufacturer’s instructions. The following primary antibodies were used: rabbit anti-caspase-3 (no. 9662), anti-caspase-9 (no. 9504), anti-cleaved caspase-9 (Asp353; no. 9509), anti-Akt (no. 4691) and anti-phospho-Akt (Ser473; no. 4060), anti-mTOR (no. 2983), anti-phospho-mTOR (Ser2448; no. 2971), anti-MKK4/7 (no. 9264), anti-phospho-MKK4/7 (Ser189/207; no. 9231), anti-phospho-JNK (Thr183/185; no. 4060), anti-CREB (no. 9197), anti-phospho-CREB (Ser133; no. 9198), anti-Bax (no. 2772), anti-Bcl-2 (no. 2870) and anti-phospho-ASK1 (Thr845; no. 3765) (all from Cell Signaling Technology, Beverly, MA, USA; 1:1,000); rabbit anti-JNK1/2/3 (T178; BS3630), anti-p-c-Jun (S63; BS4045) and anti-p-c-Jun (S73; BS4046) (all from Bioworld Technology, Minneapolis, MN, USA; 1:500–1:1,000); rabbit anti-BDNF (ab72439; 1:1,000 dilution), anti-GAP43 (ab12274; 1:500–1:2,000) and anti-NGF (ab6199; 1:1,000 dilution) (all from Abcam); mouse anti-cleaved caspase-3 (Asp175; AB3623; 1:100-1:200; Millipore); and goat anti-ASK1 (N19; sc6368; 1:200) and rabbit anti-TRAF1 (H125; sc1830; 1:200) (both from Santa Cruz Biotechnology, Santa Cruz, CA, USA). Mouse anti-GAPDH (MB001; 1:5,000-1:10,000) (Bioworld Technology) served as the internal control. Full gel scans are shown in Supplementary Fig. S14.

In vitro model of ischaemia/reperfusion

The cortical neurons were collected from Sprague–Dawley rats within 1 day of birth. Briefly, the cortices were dissociated by incubation for 15min at 37°C in 2ml 0.125% trypsin (GIBCO, Grand Island, NY, USA), followed by the addition of 4ml Dulbecco’s modified Eagle’s medium (DMEM)/F-12 (GIBCO) containing 20% fetal bovine serum (FBS, GIBCO) to inactivate the trypsin. The cell suspension was centrifuged for 10min at 1,000rpm and resuspended in DMEM containing 20% FBS. The cells were then passed through 200-μm sterile filters and seeded on plates coated with poly-L-lysine (10mgml⁻¹, Sigma). The neurons were cultured in neurobasal medium (GIBCO) fortified with B27 (GIBCO) for 24h at 37°C and 5% CO₂. AraC (10μM, Sigma) was added to the medium 24h after plating to inhibit cell proliferation, and the medium was changed every 48h. The cells were cultured for 5 days before the experiments. To model ischaemia/reperfusion conditions in vitro, the neuronal cultures were exposed to transient OGD for 60min and then returned to normal culture conditions for various periods. For OGD, the neurobasal medium was replaced with serum-free, glucose-free Locke’s buffer (154mM NaCl, 5.6mM KCl, 2.3mM CaCl₂, 1mM MgCl₂, 3.6mM NaHCO₃, 5mM HEPES and 5mgml⁻¹ gentamicin, pH 7.2), and the cultures were incubated in an experimental hypoxia chamber in a saturated atmosphere of 95% N₂ and 5% CO₂. Control cells cultured in the presence of normal levels of glucose were incubated for the same periods in a humidified atmosphere of 95% air and 5% CO₂. The neurons were incubated with anti-FasL antibody (20μgml⁻¹, Santa Cruz Biotechnology), Fas–Fc (10μgml⁻¹, R&D Systems Inc.), Z-VAD.FMK (100μM, V116, Sigma) or Necrostatin-1 (25μM; N9037; Sigma) before OGD exposure as indicated.

Construction of adenoviral vectors

Adenoviruses harbouring sequences encoding mouse TRAF1, short hairpin RNA targeting TRAF1 (shTRAF1), TRAF1 with a mutant TRAF domain (Ad-mutant), human ASK1 or dominant negative ASK1-K709R (dn-ASK1) were generated. The ORF clone of mouse TRAF1 subcloned into a pCMV6-AC-GFP shuttle vector was purchased from OriGene (MG206422). The coding sequence of human ASK1 was reverse-transcribed using the primers 5′-CCAGATTACGCTGATAGCACGGAGGCGGACGAGGG-3′ (forward) and 5′-GAAGATCTTGATTTAAGTCTGTTTGTTTCGAAAGT-3′ (reverse). dn-ASK1 was a gift from Prof. Hidenori Ichijo (The University of Tokyo, Tokyo, Japan)25. Recombinant adenoviruses were generated using an AdEasy vector kit (Stratagene, La Jolla, CA, USA). Inserts were cloned into the pShuttle-CMV vector. Plasmids were recombined with the pAdEasy backbone vector according to the manufacturer’s instructions and transfected into HEK293 cells using FuGENE transfection reagent (E2312, Roche). Recombinant adenoviruses were plaque-purified, titered to 10⁹ PFUml⁻¹, and verified by restriction digestion. Ad-mutant recombinant adenovirus harbouring TRAF1 with a mutant TRAF domain was generated using a similar protocol. To generate shTRAF1, the hairpin-forming oligonucleotides TRAF1-f, 5′-GATCCGCGTGTGTTTGAGAACATTGTTCTCGAGAACAATGTTCTCAAACACACGTTTTTTACGCGTG-3′ and TRAF1-r, 5′-AATTCACGCGTAAAAAACGTGTGTTTGAGAACATTGTTCTCGAGAACAATGTTCTCAAACACACGG-3′ were synthesized, annealed and subcloned into the shuttle vector distal to the U6 promoter. Using a similar process, recombinant adenovirus was generated. Cultured cortical neurons were transfected with adenovirus at an MOI of 100 for 48h.

Plasmid construct and transfection

To construct the EGFP-myc-TRAF1 plasmid, the mouse TRAF1 insert was amplified using PCR with the primers TRAF1-5′ and TRAF1-3′, digested with BamHI and XhoI, and ligated into pEGFP-myc-C1, which was digested with BamHI and XhoI. To generate TRAF1 fragments consisting of residues 1–261 and 261–409, EGFP-myc-TRAF1 was PCR amplified using the primer pairs TRAF1-5′ and TRAF1-N-3′ and TRAF1-C-5′ and TRAF1-3′, respectively. The PCR products were digested with BamHI and XhoI and ligated into pEGFP-myc-C1 to create an in-frame fusion with EGFP-myc. To construct recombinant Flag-cherry-ASK1, the human ASK1 insert was amplified with ASK1-5′ and ASK1-3′, digested with SalI, and ligated into pCMV-Flag-cherry. Flag-cherry-ASK1 was PCR-amplified with the primer pairs ASK1-5′ and ASK1-N1-3′, ASK1-C1-5′ and ASK1-C1-3′, and ASK1-C2-5′ and ASK1-3′ to obtain ASK1 fragments consisting of residues 1–678, 678–936 and 936–1375, respectively, and the PCR products were digested with BamHI and SalI, BamHI and SalI, or BglII and SalI, respectively. These products were ligated into pCMV-Flag-cherry digested with BamHI and XhoI to create an in-frame fusion with Flag-cherry. All plasmids were verified through sequencing. Plasmid DNA was transfected into cells at 50% confluence in 12-well plates (Greiner Bio One, Monroe, NC, USA) with FuGENE transfection reagent (E2312, Roche) for 48h, according to the manufacturer’s instructions.

Immunoprecipitation

Cultured HEK293T cells were co-transfected with pEGFP-myc-TRAF1 and pFlag-cherry-ASK1 for 48h with FuGENE transfection reagent (Roche). Primary cortical neurons were subjected to OGD exposure with or without Ad-TRAF1 infection, as indicated. The cells were lysed in IP buffer (20mM Tris-HCl, pH 8.0, 100mM NaCl, 1mM EDTA, and 0.5% NP-40) containing protease inhibitor cocktail (Roche). After 20min of incubation at 4°C with gentle vortex mixing, the cell lysates were centrifuged at 13,000 × g for 10min at 4°C. Sample aliquots (500μl) were precleared with 10μl Protein A/G-agarose beads (11719394001, 11719386001, Roche) and incubated with 1μg of antibody or control IgG overnight at 4°C, according to the manufacturer’s recommendations. The beads were washed 5–6 times with 800μl cold IP buffer, resuspended in 1 × loading buffer, and heated at 95°C for 5min. The cell lysates and immunoprecipitates were analysed through western blotting, as described above.

GST pull-down assay

GST-TRAF1 was constructed using pGEX-4T-1 and expressed in E. coli Rosetta (DE3) cells. The cultures were induced with IPTG; 10ml aliquots of cells were harvested, and fusion proteins were purified using glutathione-Sepharose 4B beads (GE Healthcare, Chalfont St Giles, UK), according to the manufacturer’s instructions. The lysates from Flag-cherry-ASK1-transfected HEK293T cells were added to the beads and incubated for 4h at 4°C in IP buffer supplemented with protease inhibitor cocktail. The GST tag served as a negative control. The beads were extensively washed with IP buffer and analysed by western blotting.

Confocal microscopy

HEK293T cells were seeded on coverslips in 24-well plates. After co-transfection of pFlag-cherry-ASK1 and pEGFP-myc-TRAF1 for 48h, the cells were fixed in 4% fresh paraformaldehyde for 15min, permeabilized with 0.2% Triton X-100 in PBS for 5min and incubated in Image-IT FX signal enhancer (I36933, Invitrogen) for 30min. The cells were washed three times with TBST and stained with 1gml⁻¹ DAPI for 15min. The slides were mounted with mounting solution (D2522, Sigma), and images were obtained with a confocal laser-scanning microscope (Fluoview 1000, Olympus).

Analysis of cell viability and LDH release

Cell viability was evaluated using a non-radioactive cell counting kit-8 (CCK-8) assay (CK04, Dojindo, Kumamoto, Japan), according to the manufacturer’s instructions. LDH release was analysed using a colorimetric LDH cytotoxicity assay (G1782, Promega, Madison, WI). Three independent experiments were performed.

The authors thank Professor Hidenori Ichijo, who provided the kind gift of dominant-negative ASK1 cDNA. This work was supported by grants from the National Natural Science Foundation of China (no. 81100230, no. 81330005 and no. 81070089), National Science and Technology Support Project (no. 2013YQ030923-05, no. 2011BAI15B02 and no. 2012BAI39B05), and the National Basic Research Program of China (no. 2011CB503902).