Introduction

A transient interruption of the blood supply to the brain triggers a cascade of events culminating in selective neuronal death in vulnerable regions, such as hippocampal CA1 pyramidal cells, Purkinje cells of the cerebellum, striatal medium spiny neurons, and neurons in the third to fifth layers of the cerebral cortex [26, 46]. Morphological studies have shown that these neurons die, in varying proportions, from necrosis and/or apoptosis. The relative contribution of each pathway is dependent on the duration of ischemia and other as yet unidentified factors (reviewed in [34]). Moreover, both pathways may be active simultaneously. The precise events leading to neuronal death are under intense investigation [66].

Upon brain reperfusion following ischemia, there is widespread inhibition of neuronal protein synthesis that persists in selectively vulnerable neurons (SVNs) destined to die (reviewed in [15]). The immediate inhibition of ribosomal translation initiation is due to inhibited delivery of the initiator methionine caused by a greater than 20-fold increase in phosphorylation of serine 51 on eukaryotic initiation factor 2α [eIF2α, phosphorylated form eIF2α(P)] [15]. Several investigators have shown a link between apoptosis and phosphorylation of eIF2α [1, 9, 17, 56]. In vitro studies have shown that expression of mutant eIF2α (S51D) mimicking eIF2α(P) induces apoptosis, and expression of non-phosphorylatable eIF2α (S51A) blocks induction of apoptosis [17, 56]. The nature of the relationship of eIF2α(P) to apoptosis is unknown.

One early event in initiating apoptosis is the release of cytochrome c from the intramembranous space of mitochondria. This allows formation of the apoptosome complex, consisting of cytochrome c, apoptosis-activating factor 1 (Apaf-1), and caspase-9, which proteolytically activates caspase-3, a key apoptosis executioner (reviewed in [63]). Indeed, after 3–4 h of reperfusion, many SVNs begin to show release of cytochrome c and caspase-9 from mitochondria into the cytoplasm [7, 58]. In this study, we explored the potential relationship between eIF2α(P) and cytochrome c release utilizing immunohistochemistry. We found virtually complete co-mapping between eIF2α(P) and cytochrome c in CA1 pyramidal neurons at 4 h of reperfusion following 10-minute cardiac arrest. In addition, eIF2α(P) appears before cytochrome c release, suggesting that eIF2α(P) is upstream of cytochrome c release.

Materials and methods

Animal model

All animal experiments were approved by our institutional review board and were conducted following the Guide for the Care and Use of Laboratory Animals (National Research Council, revised 1996). We utilized our well-characterized rat model of normothermic 10-min cardiac arrest followed by resuscitation, which causes: (1) neurological functional deficits in long-term survivors [55], (2) substantial loss of hippocampal SVNs [49], (3) inhibited translation initiation [11] associated with eIF2α phosphorylation [12] and μ-calpain-mediated degradation of eIF4G [12, 38], and (4) lipid peroxidation in SVNs [65] with associated ultrastructural alterations of the endoplasmic reticulum (ER) and Golgi apparatus [47]. Male Long Evans rats weighing 300–400 g were anesthetized with ketamine (70 mg/kg intraperitoneally) and xylazine (5 mg/kg intraperitoneally). Rats were orotracheally intubated and mechanically ventilated with room air at 80 breaths per minute, tidal volume of 10 ml/kg, and a positive end expiratory pressure of 3 cm H20. Femoral artery and femoral vein catheters were inserted and arterial pressure was continuously monitored. Core body temperature was maintained with a servo-controlled heating pad at 37.0±0.5°C for the entire experimental period. The electrocardiogram was continuously monitored by limb leads. After instrumentation, circulatory arrest was induced by thoracic compression [16, 64] and was confirmed by the complete loss of arterial pressure, which occurred within 5 s. After a 10-min cardiac arrest period, resuscitation was initiated by chest compressions (200–300/min) and mechanical ventilation (tidal volume 1 ml/100 g, 80 breaths per minute, 3 cm positive-end-expiratory pressure) with supplemental O2. The rate and depth of chest compressions was adjusted to maximize the arterial diastolic pressure. Epinephrine (10 µg/kg IV) was given during resuscitation. ECG and arterial pressure tracings were evaluated for return of spontaneous circulation, defined as a systolic blood pressure greater than 60 mmHg. If return of spontaneous circulation (ROSC) did not occur after 3.5 min of resuscitation efforts, the resuscitation was abandoned and the animal was not used in any studies. The time from onset of ischemia to ROSC was 700±100 s. In the early post-resuscitation phase, systolic blood pressure was maintained above 80 mmHg using 0.9% saline and dopamine infusion at 10–20 μg/kg/min as necessary. Artificial ventilation was continued until the end of the experiment.

Western blot analysis

The brain was homogenized and postmitochondrial supernatants were prepared as previously described [12, 13, 30, 38]. Proteins (50 μg) were separated by electrophoresis in SDS-polyacrylamide gel and electrophoretically transferred to nitrocellulose membranes. Cytochrome c and eIF2α(P) were detected with a mouse primary monoclonal antibody (BD PharMingen, San Diego, CA) and a rabbit primary polyclonal antibody (BioSource International, Camarillo, CA), respectively, and visualized by the enhanced chemiluminescence procedure (Amersham Pharmacia Biotech).

Fluorescence immunohistochemistry

Transaortic perfusion fixation with 50% methanol:10% acetic acid was performed and 40-μm vibratome sections prepared as previously described [13]. This fixative was used in preference to aldehydes because it allows much better anti-eIF2α(P) antibody penetration without resorting to antigen recruitment procedures [18]. Sections were quenched in 3% hydrogen peroxide (H2O2) for 10 min, washed three times with TRIS-buffered saline (TBS), solubilized in 0.1% Triton X-100 in TBS for 10 min, washed three times, and blocked in 10% normal goat serum in TBS for 1 h. Sections were then incubated in primary antibody (1:100 in TBS) at room temperature overnight, and unbound antibody then washed off with four TBS rinses. Fluorescent secondary antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA) anti-rabbit-FC rhodamine red-X (RRX) and anti-mouse-FC Cy2 were used to visualize eIF2α(P) and cytochrome c, respectively. Sections were incubated in the appropriate secondary antibody (1:100 in TBS) at room temperature overnight. Following four 15-min TBS rinses, the sections were mounted and coverslipped. Fluorescence photomicrographs were obtained utilizing a Nikon digital camera on a Leica DML fluorescence microscope with appropriate incident wavelengths (492 nm for Cy2, and 570 nm for RRX). For double immunofluorescence, sections were processed as above except that single sections were incubated first with the anti-cytochrome c primary and secondary antibodies and subsequently with the anti-eIF2α(P) primary and secondary antibodies. The resulting images were overlaid utilizing Adobe Photoshop.

Results

Because we were interested in the potential relationship between persistence of eIF2α(P) and neuronal selective vulnerability, we first utilized Western blotting to examine relative levels of eIF2α(P) by region in brain homogenates obtained from three rats in each of four experimental groups: (1) non-ischemic controls, (2) 10-min ischemia followed by no reperfusion, (3) 10-min reperfusion, and (4) 60-min reperfusion. There was no increase of eIF2α(P) during ischemia (Fig. 1), but after 10 min of reperfusion a whole-brain homogenate showed a 24-fold increase over controls. Similarly, brain homogenates from microdissected regions of the brain stem, cerebellum, cortex, hippocampus, and thalamus all showed large increases. By 1 h of reperfusion, there was a consistent trend toward reduction of the eIF2α(P) levels in all regions except the cortex and hippocampus, although this decrease was statistically significant only in the ischemia-resistant brain stem.

Fig. 1.
figure 1

Regional changes in eIF2α(P) Western blot density during the first hour of reperfusion with the data normalized against the NIC mean. Groups are: (1) non-ischemic controls (NIC) and (2) 10 min. ischemia followed by no reperfusion (10I), (3) 10 min. reperfusion (10R), and (4) 60 min. reperfusion (60R). ANOVA followed by Fisher's LSD post hoc tests showed that all reperfused samples had significantly greater eIF2α (P) levels (*P<0.05) than seen in NIC or 10I; NIC and 10I were not different for any group. The decrease between 10 and 60 min reperfusion in the brain stem was also significant [eIF2α(P) phosphorylated eukaryotic initiation factor 2α, BS brain stem, CER cerebellum, CTX cortex, HIP hippocampus, THAL thalamus, WB whole-brain homogenate]

Full size image

We chose the 4-h reperfusion time point to investigate potential colocalization of eIF2α(P) and cytoplasmic cytochrome c because of the evidence from Cao et al. [7] that mitochondrial release of cytochrome c had begun by that time. By Western blot we observed that levels of eIF2α(P) remained approximately sixfold elevated after 4-h reperfusion in forebrain (rostral to the thalamus) homogenates (Fig. 2A). Immunoblot examination of the anti-eIF2α(P) (Fig. 2B) and anti-cytochrome c (Fig. 2C) antibodies showed that both were substantially monospecific.

Fig. 2.
figure 2

A eIF2α (P) after 10-min and 4-h reperfusion is increased ~24- and 6-fold, respectively, over controls in forebrain homogenates. B The rabbit polyclonal anti-eIF2α (P) is monospecific for eIF2α(P) produced by 30-min thapsigargin (tg) treatment of NB104 cells, and generates little signal in homogenates from normal control brains; it is monospecific for eIF2α(P) in brain homogenates obtained after 10-min cardiac arrest and 10-min reperfusion. C The mouse monoclonal anti-cytochrome c is monospecific and recognizes much more cytochrome c in the mitochondrial pellet than in post-mitochondrial supernatant (PMS) obtained from normal rat brains

Full size image

Immunofluorescence examination of eIF2α(P) and cytochrome c was performed on three animals each in two groups: (1) control and (2) 10-min cardiac arrest with 4-h reperfusion. Representative photomicrographs are shown. Omission of primary antibody ablated the strong immunofluorescence signal seen in reperfused brains (not shown). Control brains revealed only background fluorescence for eIF2α(P) and scant punctate cytoplasmic immunofluorescence for cytochrome c (shown for hippocampus CA1 in Fig. 3). Although cytochrome c is constitutively present in the mitochondrial intramembranous space, other investigators have also noted scant immunostaining before cytochrome c is released into the cytoplasm [32, 40].

Fig. 3.
figure 3

Representative photomicrographs of immunofluorescence staining for eIF2α(P) or cytochrome c in the CA1 region (arrowheads indicate CA1 border) from normal controls show little signal above background, although sparse punctate cytoplasmic immunofluorescence for cytochrome c is noted. Results were confirmed by at least three independent studies for each time point

Full size image

In marked contrast to controls, there was a strong immunofluorescence signal in corresponding populations of SVNs in the striatum and hippocampal CA1 and hilus after 4-h reperfusion when serial consecutive sections were stained for either eIF2α(P) or cytochrome c (Fig. 4). There was a well-demarcated reduction of fluorescence for both species in the injury-resistant CA2 (left of arrow in Fig. 4A, B) as compared to CA1. The injury-resistant hippocampal dentate granule cells also showed a paucity of fluorescence for either species, although a few cells in the polymorphic layer nearest the hilar cells demonstrated fluorescence (Fig. 4C, D). In the striatum, the radial bundles did not fluoresce, but the vulnerable medium spiny neurons showed fluorescence for both eIF2α(P) and cytochrome c (Fig. 4E, F).

Fig. 4.
figure 4

After 10-min cardiac arrest and 4-h reperfusion, representative immunofluorescence photomicrographs for eIF2α(P) (red) or cytochrome c (green) in serial consecutive sections are shown for: the junction (large arrows) of the hippocampus CA1 and CA2 (A, B); the dorsal and ventral blades of the DG and the interposed hilus (C, D); and the striatum (E, F). A striatal radial bundle is outlined in E and F. Results were confirmed by at least three independent studies for each time point (DG dentate granule cells)

Full size image

In the cortex after 4 h of reperfusion, we observed a somewhat attenuated eIF2α(P) immunofluorescence that was mainly present in layer III pyramidal neurons without any consistent cytoplasmic cytochrome c immunofluorescence in these cells (Fig. 5).

Fig. 5.
figure 5

After 10-min cardiac arrest and 4-h reperfusion, representative immunofluorescence photomicrographs for eIF2α (P) (red) or cytochrome c (green) in serial consecutive sections are shown for the cerebral cortex. Many layer III neurons show some degree of cytoplasmic immunofluorescence for eIF2α (P) (left); however, the lower magnification image at the right (small arrow points to the pial surface) shows the widespread paucity of immunofluorescence for cytoplasmic cytochrome c in this region. Results were confirmed by at least three independent studies for each time point

Full size image

Double immunofluorescence studies in the three areas shown in Fig. 4 revealed strong immunofluorescence signals for both eIF2α(P) and cytoplasmic cytochrome c with nearly complete cellular congruity (CA1 shown in Fig. 6). Higher magnification shows that cytochrome c staining in normal controls (Fig. 7, left) is mainly punctate as would be expected with mitochondrial localization. However, after 4-h reperfusion following 10-min cardiac arrest, there is diffuse cytoplasmic cytochrome c staining consistent with its release into the cytosol (Fig. 7, right).

Fig. 6.
figure 6

Representative double immunofluorescence photomicrographs for eIF2α(P) and cytochrome c in the hippocampus CA1 at 4-h reperfusion after 10-min cardiac arrest. Results were confirmed by at least three independent studies for each time point

Full size image
Fig. 7.
figure 7

Representative immunofluorescence photomicrographs for cytochrome c in the hippocampus CA1 from brain perfusion fixed with 4% paraformaldehyde. In normal controls (left), there is mainly punctate staining. After 4-h reperfusion following 10-min cardiac arrest (right), there is diffuse cytoplasmic staining consistent with cytochrome c release into the cytosol. Bar 10 μm

Full size image

Discussion

Mitochondria are thought to play a pivotal role in the early phase of apoptosis in mammalian cells [54, 63]. Following certain apoptotic stimuli, cytochrome c is released from the mitochondria and binds to Apaf-1, which leads to activation of caspase-9 and other caspases, including caspase-3. Caspase-3, believed be the final committed step in neuronal apoptosis, cleaves several cellular targets producing the hallmark morphological features of apoptosis. Sugawara et al. [58] have shown that mitochondrial release of cytochrome c corresponds to the selective vulnerability of hippocampal CA1 neurons in rats after transient global cerebral ischemia. At present, the apoptotic signaling pathways leading to mitochondrial permeability are not well defined. There are several pathways by which eIF2α(P) and apoptosis might be linked. Further work is needed to discern which pathway is correct, and it is certainly possible that multiple pathways may be simultaneously active [29].

A direct mechanism from persistent eIF2α phosphorylation to apoptosis induction could involve an eIF2α(P)-induced bypass-scanning mechanism of translation initiation for ATF4 [20] (the transcription factor for CHOP) and CHOP [51]. CHOP (C/EBP homology protein) production is required for the loss of mitochondrial membrane integrity and apoptosis caused by depletion of ER Ca2+ following nitric oxide-induced nitrosylation and inactivation of the ER Ca2+ pump [42], and its synthesis precedes release of cytochrome c to the cytoplasm [61]. ER stress-induced CHOP transcription is blocked in cells containing a non-phosphorylatable mutant of eIF2α (S51A) [51]. Although Paschen et al. [45] have shown that CHOP transcription is induced during brain reperfusion, CHOP protein levels in the reperfused brain have yet to be determined.

A second pathway by which eIF2α(P) and apoptosis might be linked is through caspase-8. Caspase-8 can cleave the eIF2α kinase PKR (RNA-dependent protein kinase), which separates the PKR regulatory domain from its kinase domain; the released kinase domain can then efficiently phosphorylate eIF2α [50]. Caspase-8 also cleaves Bid, and the resulting Bid fragment interacts with the mitochondrial membrane to release cytochrome c [33, 35] by a process that does not involve permeability transition [52]. Such a mechanism for cytochrome c release is attractive because Fiskum's group has developed compelling evidence that brain ischemia and reperfusion do not cause mitochondrial permeability transition [3]. Yin et al. [68] have recently shown activated Bid and early release of cytochrome c in the brain following transient cerebral ischemia. However, previous work in our laboratory has shown that mice bearing a homozygous PKR knockout do not have reduced levels of eIF2α(P) during early reperfusion, which argues against this pathway [14].

A third potential pathway involves the role of reactive oxygen species in the phosphorylation of eIF2α and release of cytochrome c. Neurons produce large amounts of H2O2 following ischemia/reperfusion [24], and H2O2 inhibits protein translation in cortical neurons by a process that involves the phosphorylation of both eIF2α and eEF2 [2]; the relative contribution of these two events depends on the duration of H2O2 treatment. H2O2 treatment is known to induce release of cytochrome c from cultured cells [57, 60, 67].

Finally, ischemia/reperfusion-induced ER stress could trigger a cytochrome c-independent apoptosis pathway involving IRE1α and caspase-12. The reperfusion-induced phosphorylation of eIF2α is likely caused by activation of the eIF2α kinase PERK (PKR-like ER kinase); knockouts of the other three eIF2α kinases (PKR, HRI, GCN2) do not reduce levels of eIF2α(P) during reperfusion [30]. The only known mechanism by which PERK is activated is as a component of the ER unfolded protein response (UPR) [19, 25]. Originally described as a response to an overload of unfolded proteins in the ER [28], several types of ER stress trigger a compensatory response, the UPR, whose most proximal sensors are the ER membrane-bound proteins ATF6, IRE1α, and PERK. Thus, induction of the UPR is also expected to cause activation of IRE1α, which is both a kinase and a RNase [53].

In mammalian cells active IRE1α carries out cytoplasmic endoribonucleolytic processing of XBP1 mRNA [6]. Translation of the processed XBP1 participates in the resolution of the UPR [31] by generating a transcription factor that strongly enhances gene transcription, such as that of the ER chaperone GRP78. Resolution of the ER stress is associated with GRP78 binding of the ER luminal domains of both IRE1α and PERK, which attenuates their activation [5]. In the case of cultured hippocampal neurons, antisense-mediated depletion of GRP78 substantially enhances cell death induced by glutamate or Fe2+-mediated radical damage [70]. Paschen's group has preliminary evidence of XBP1 processing consistent with IRE1α activation during post-ischemic reperfusion in a stroke model (personal communication), and induction of GRP78 transcription by brain ischemia and reperfusion is well established [15].

Activated IRE1 recruits tumor necrosis factor receptor-associated factor-2 (TRAF2) [62]. Normally TRAF2 forms a stable complex with procaspase-12 [69], but ER stress [37] dissociates TRAF2 from procaspase-12, which then oligomerizes, is activated [69] and can cleave caspase-9 [36, 48], which is presumably followed by activation of caspase-3 [8]. In addition, activation of IRE1α also leads to activation of the apoptosis-associated stress-induced protein Ser/Thr kinases known as SAPKs (stress activated protein kinases) or JNKs (Jun N-terminal kinases) via interaction of the IRE1α cytoplasmic domain with TRAF2 [62] and JIK (Jun inhibitor kinase) [69]. During early reperfusion there is rapid activation of JNK1 in SVNs [59] associated with enhanced N-terminal Jun phosphorylation [21].

The initial roles of both PERK and IRE1α in the UPR appear protective: PERK by inhibiting protein synthesis to reduce the load of new peptides requiring ER luminal processing; and IRE1α by up-regulation of genes encoding chaperones that facilitate the protein-folding process in the ER. The ischemia-reperfusion-induced activation of PERK [30] suggests that ER stress has caused activation of the UPR, consistent with evidence developed by Paschen [43, 44].

Importantly, if the ER stress is not resolved, both PERK and IRE1α can also drive apoptotic mechanisms [39], and there is evidence of prolonged ER stress in the reperfused brain. Brain ischemia and reperfusion induce abnormal protein aggregates in the ER and other neuronal compartments [22, 23]; these aggregates clear from injury-resistant neurons but persist in vulnerable neurons that progress to cell death. Moreover, Kohno et al. [27] have shown prolonged loss of ER luminal Ca2+ in reperfused SVNs, and low ER Ca2+ concentrations leads to inhibition of the ER luminal protein disulfide isomerases (PDI) and thus causes misfolding of newly synthesized peptides in the ER lumen [4, 10, 41].

In this study, we have shown that there are large increases in eIF2α(P) in both vulnerable and non-vulnerable brain regions at 10 min of reperfusion following 10-min cardiac arrest. By 1 h of reperfusion, there was a consistent trend toward reduction of the eIF2α(P) levels in all regions except the cortex and hippocampus. We also demonstrated by double immunofluorescence that eIF2α(P) and cytochrome c show substantial congruity at 4-h reperfusion; we did not observe cytoplasmic cytochrome c in any neurons that did not also display persistent eIF2α(P) immunostaining. These findings suggest that there is prolonged activation of the UPR in the reperfused brain. These results are also consistent with a hypothesis that persistent eIF2α(P) reflects injury mechanisms that are causally upstream of release of cytochrome c to the cytoplasm, an event documented as proapoptotic [66].

Interestingly, at 4-h reperfusion, persistent eIF2α(P) is seen in layer III cortical neurons without cytoplasmic release of cytochrome c, while neurons in the other selectively vulnerable areas demonstrate both persistent eIF2α(P) and cytoplasmic release of cytochrome c. Because cortical pyramidal neurons are more tolerant of ischemia than hippocampal CA1 pyramidal neurons, this suggests that mechanisms other than simply a transient elevation in eIF2α(P) levels play an important role in determining the ultimate fate of the post-ischemic neuron.