Mitochondrial BNIP3 upregulation precedes endonuclease G translocation in hippocampal neuronal death following oxygen-glucose deprivation

Background Caspase-independent apoptotic pathways are suggested as a mechanism for the delayed neuronal death following ischemic insult. However, the underlying signalling mechanisms are largely unknown. Recent studies imply the involvement of several mitochondrial proteins, including endonuclease G (EndoG) and Bcl-2/adenovirus E1B 19 kDa-interacting protein (BNIP3), in the pathway of non-neuronal cells. Results In this report, using western blot analysis and immunocytochemistry, we found that EndoG upregulates and translocates from mitochondria to nucleus in a time-dependent manner in cultured hippocampal neurons following oxygen-glucose deprivation (OGD). Moreover, the translocation of EndoG occurs hours before the observable nuclear pyknosis. Importantly, the mitochondrial upregulation of BNIP3 precedes the translocation of EndoG. Forced expression of BNIP3 increases the nuclear translocation of EndoG and neuronal death while knockdown of BNIP3 decreases the OGD-induced nuclear translocation of EndoG and neuronal death. Conclusion These results suggest that BNIP3 and EndoG play important roles in hippocampal neuronal apoptosis following ischemia, and mitochondrial BNIP3 is a signal protein upstream of EndoG that can induce neuronal death.


Background
The hippocampus, a very important structure for learning and memory, is among the most vulnerable brain regions after global cerebral ischemia. The rapid decrease of oxygen and glucose in the ischemic region can trigger delayed neuronal death [1], and reperfusion may further exacerbate the injury. The delayed cell death occurs primarily through an apoptotic mechanism [2]. Evidence has accumulated that a large proportion of the delayed neuronal death is mediated by caspase-independent pathways [3][4][5]. However, the signalling mechanisms remain largely unclear. Recent studies have implicated mitochondrial proteins, such as endonuclease G (EndoG) and Bcl-2/adenovirus E1B 19 kDa-interacting protein (BNIP3), as players involved in this pathway in non-neuronal cells [6].
EndoG is an endonuclease normally localized in the mitochondrial intermembrane space, and translocates to the nucleus when cells are exposed to an apoptosis-inducing stimulus. After moving to the nucleus, EndoG cleaves chromatin DNA into nucleosomal fragments independent of caspases [7]. The compartmentalization of mitochondria plays the major role in EndoG trafficking. The EndoG location site might indicate that this enzyme is not an instrument for immediate response to cell injury [8].
BNIP3 is a member of a unique family of death-inducing mitochondrial proteins [9]. In neurons, the expression of BNIP3 is undetectable under normal conditions [10] but can be induced by hypoxia/ischemia and oxidative stress [11,12]. The BNIP3-induced non-neuronal cell death is characterized by mitochondrial damage but is independent of caspase activation and cytochrome c release [10]. It is presently unclear how BNIP3 initiates neuronal death.
In this study, we profiled BNIP3 and EndoG expression and translocation in cultured hippocampal neurons following oxygen-glucose deprivation (OGD). We demonstrate here that EndoG upregulates and translocates from mitochondria to nucleus in a time-dependent manner. Moreover, the translocation of EndoG occurs hours before the observable nuclear pyknosis. Importantly, we investigated the causal relationship between mitochondrial BNIP3 upregulation and EndoG translocation and nuclear pyknosis by over expressing or knocking down of BNIP3. Our findings strongly support a role for BNIP3 as a signal protein upstream of EndoG leading to the induction of neuronal death.

OGD induces EndoG upregulation in primary neuronal cultures
Primary rat hippocampal neurons at day 12 in vitro were subjected to OGD for 4 h and reoxygenation for 0 h, 2 h, 6 h, 12 h and 24 h. Normal cultured control and EBSStreated neurons under normoxic condition were set as controls. Levels of EndoG expression were determined by western blot analysis. EndoG immunoreactivity was identified as the ~35 kDa protein band in the whole cell fraction after treatment. Immunoblotting with a monoclonal β-actin antibody was performed as standards for calculation of EndoG expression. As shown in Figure 1, total amount of EndoG did not vary until reoxygenation was prolonged to 6 h and accumulated to the highest level at 12 h of reoxygenation.

EndoG translocates from the mitochondria to the nucleus after reoxygenation
We next examined the subcellular localization of EndoG to determine whether OGD could induce the translocation of EndoG. Mitochondrial and nuclear fractions were prepared by differential centrifugation. Cox IV antibody for mitochondrial and histone H3 antibody for nuclear fractions were used as sample loading controls. EndoG level in the mitochondrial fraction began to decrease significantly at 2 h of reoxygenation and further decreased with time exposed to reoxygenation (Figure 2A and 2B). Meanwhile, EndoG in nuclear fraction showed a significant increase at the corresponding time point and accumulated with increased periods of reoxygenation ( Figure  2C and 2D). These data suggest that the released EndoG from mitochondria translocates to nucleus.

Nuclear EndoG translocation precedes OGD-induced neuronal death
To evaluate the role of nuclear EndoG translocation in OGD, we investigated the relation between EndoG translocation and cell death. At different time points after reox-Upregulation of EndoG expression in reoxygenated hippoc-ampal neuronal cultures following OGD ygenation, the intracellular localization of EndoG was detected by immunocytochemistry; the morphologically damaged neurons were identified by condensed and fragmented nuclei using DAPI nuclear staining. The result revealed that EndoG was mainly located in the intracellular space out of the nucleus shown as red fluorescence in normal control group or in EBSS-treated group under normoxic condition ( Figure 3A and 3B). Consistent with the results from the previous western blot analysis, the nuclear translocation of EndoG started at 2 h after reoxygenation and increased progressively ( Figure 3C-E). Significant numbers of neurons displaying EndoG nuclear staining were observed 6 h after reoxygenation, a time point when no obvious morphological neuronal damage could be observed ( Figure 3C and 3E). Thereafter, although the number of morphologically damaged neurons also increased with time, a great quantity of pyknotic nuclei could not be observed until 12 h after reoxygenation. Furthermore, the proportions of morphologically damaged neurons were less than that of the EndoG translocated neurons at 12 h and 24 h, respectively ( Figure 3D and 3E). These results demonstrate that EndoG translocation from mitochondria to nucleus precedes neuronal cell death.

BNIP3 mediates OGD-induced hippocampal neuronal death
To explore the role of BNIP3 in OGD-induced hippocampal neuronal cell death, we first observed the expression of BNIP3 following reoxygenation. Western blot analysis revealed that BNIP3 bands with molecular weights of 30 kD (monomer) and 60 kD (dimer) started to increase at 2 h of reoxygenation ( Figure 4A and 4B). We next examined whether the expression of BNIP3 was sufficient to induce neuronal death and also if BNIP3 was necessary to mediate neuronal death after OGD. Hippocampal neurons were transfected with pEGFP-C3-rBNIP3 plasmid or pEGFP-C3 plasmid as a control. As shown in Figure 4C, exogenous BNIP3 expression in normal cultured neurons led to about 40% of cell death after transfection for 24 h. To knockdown the expression of BNIP3, BNIP3-miRNA construct was generated and co-transfected with pEGFP-Translocation of EndoG from mitochondria to nuclei following reoxygenation C3-rBNIP3 plasmid encoding the full length of rat BNIP3 in HEK293 cells. The inhibition efficiency of BNIP3-miRNA for BNIP3 expression was 80% compared with the neg-miRNA transfected controls as identified by western blot analysis ( Figure 4D and 4E). Addition of BNIP3-miRNA to hippocampal neurons significantly reduced OGD-induced nuclear pyknosis when detected 12 and 24 h after reoxygenation ( Figure 4F and 4G).

BNIP3 initiates the EndoG translocation
Finally, we tested the hypothesis that BNIP3 initiates the translocation of EndoG to nucleus in OGD-induced neuronal death. Because BNIP3 predominantly localizes to the mitochondria and homodimerization appears to be a feature of mitochondrial localization and important for BNIP3's function [9], we explored the subcellular localization of OGD-induced BNIP3. Immediately after reoxygenation, upregulation of mitochondrial BNIP3 (dimer) was detected, 2 h earlier than the EndoG translocation ( Figure  5A and 5B). We next examined whether forced expression of BNIP3 induced EndoG translocation. As shown in Figure 5C and 5D, the proportion of nuclear EndoG positive staining neurons increased significantly in exogenous BNIP3 expressing neurons compared to that in pEGFP-C3 transfected control neurons. Furthermore, we explored whether knockdown of BNIP3 could reduce the OGDinduced nuclear translocation of EndoG. Based on the earlier result that EndoG began to significantly translocate to nucleus at 2 h after reoxygenation, we wanted to observe the effect that BNIP3 inhibition would have on the percentage of neurons with nuclear EndoG at 2 h and 6 h after reoxygenation, respectively. Knockdown of BNIP3 expression significantly reduced EndoG nuclear translocation ( Figure 5E and 5F).

Recent biochemical and genetic studies have revealed that
EndoG is an important mitochondrial protein that emanates from the mitochondria during apoptosis and facilitates degradation of nuclear chromatin [7,13]. The present study, for the first time, strongly implies an essential role of EndoG in post-ischemic hippocampal neuronal cell death in vitro. Western blot analysis and immunocytochemistry demonstrate that EndoG not only upregulates in hippocampal neurons following reoxygenation but also translocates from mitochondria to nucleus. This finding is consistent with previous results on other ischemic models, including cortical neuronal cultures subjected to hypoxia [11], and cortical neurons of mice subjected to transient or permanent focal cerebral ischemia [14,15]. Using subcellular fractionation with a detailed time course, we confirmed that EndoG translocation markedly precedes the appearance of biochemical markers of cell death. The good temporal and spatial relationship between EndoG translocation and nuclear pyknosis suggests a causal role of EndoG translocation in OGDinduced hippocampal neuronal death.
The present study also identifies BNIP3 and EndoG translocation as early events in hippocampal neurons subjected to OGD. The increase in mitochondrial BNIP3 expression was observed immediately after reoxygenation. The translocation of EndoG was detectable at 2 h EndoG nuclear translocation preceded neuronal cell death following reoxygenation Figure 3 EndoG nuclear translocation preceded neuronal cell death following reoxygenation. (A-D) Confocal laser scanning microscope images of EndoG immunoreactivity (red) at control, EBSS, OGD/R6 h, OGD/R24 h, respectively. Co-staining with DAPI (blue) allowed the identification of nuclear translocation of EndoG. The white arrow indicates the nucleus with EndoG translocation but without pyknosis, and the white arrowhead indicates the pyknotic nucleus with EndoG translocation; (E) Percentage of damaged neurons and neurons displaying nuclear EndoG at different time points following reoxygenation. *p < 0.05 vs. control group, n = 9.
and increased significantly at 6 h after reoxygenation. This is consistent with a previous report in a mouse model of brain ischemia, in which EndoG translocation was observed 4 h after transient focal cerebral ischemia [14]. Although in a previous study we found that BNIP3 and EndoG translocation occurred relatively late in cultured cortical neurons subjected to hypoxia [11], the mitochondrial translocation of the BNIP3 was before the nuclear translocation of the EndoG in both studies, which strongly supports that BNIP3 acts as an upstream signal of EndoG. The discrepancy might result from either different neuron types examined or different ischemic models used. Compared with hypoxia, OGD is a more severe stress to neurons and re-supply of oxygen and glucose may further exacerbate the injury. In fact, we detected an upregulation of BNIP3 in mitochondria at 0 h after reox-BNIP3 is a mediator of OGD-induced hippocampal neuronal death ygenation, which might imply that the mitochondrial translocation of BNIP3 occurred and the signalling cascade triggered by BNIP3 started during OGD. The expression of BNIP3 in mitochondria further increased with prolonged reoxygenation, and knockdown of BNIP3 reduced cell death, supporting a role of BNIP3 in activating the cell death program. We also found that forced expression of BNIP3 increased nuclear translocation of EndoG. On the other hand, knockdown of BNIP3 expression reduced OGD-induced EndoG translocation. Therefore, the present findings argue for the importance of mitochondrial BNIP3 upregulation as an upstream modulator of EndoG translocation in ischemic neuronal injury.

Conclusion
In the present study we provide the first detailed description of the time-dependent subcellular localization of EndoG and BNIP3 in the cultured hippocampal neurons following OGD. By exploring the temporal relationship between the neuronal nuclear pyknosis and the translocation of the mitochondrial death-related proteins, BNIP3 and EndoG, we have been able to suggest an important role of mitochondrial BNIP3 upregulation and EndoG translocation in neuronal death. Combined with the findings that forced expression of BNIP3 increases EndoG translocation and neuronal death, and knockdown of BNIP3 decreases EndoG translocation and neuronal death, our results support the role of mitochondrial BNIP3 as a signal protein upstream of EndoG leading to the induction of neuronal death.

Animals
Neonatal 1 d Sprague-Dawley rats were provided by Southern Medical University and conditions regarding health and hygiene were confirmed. All experimental procedures in this study were performed within National Institutes of Health guidelines for the care and use of laboratory animals.