Address correspondence and reprint requests to David C. Henshall, Department of Physiology & Medical Physics, Royal College of Surgeons in Ireland, 123 St. Stephen's Green, Dublin 2, Ireland. E-mail: email@example.com
FOXO3a is member of the Forkhead box class O transcription factors, which functions in diverse pathways to regulate cellular metabolism, differentiation, and apoptosis. FOXO3a shuttles between the cytoplasm and nucleus and may be activated in neurons by stressors, including seizures. A subset of nuclear transcription factors may localize to mitochondria, but whether FOXO3a is present within brain mitochondria is unknown. Here, we report that purified mitochondrial fractions from rat, mouse, and human hippocampus, as well as HT22 hippocampal cells, contain FOXO3a protein. Immunogold electron microscopy supported the presence of FOXO3a within brain mitochondria, and chromatin immunoprecipitation analysis suggested FOXO3a was associated with mitochondrial DNA. Over-expression of a mitochondrially targeted FOXO3a fusion protein in HT22 cells, but not primary hippocampal neurons, conferred superior protection against glutamate toxicity than FOXO3a alone. Mitochondrial FOXO3a levels were reduced in the damaged region of the mouse hippocampus after status epilepticus, while mitochondrial fractions from the hippocampus of patients with temporal lobe epilepsy displayed higher levels of FOXO3a than controls. These results support mitochondria as a site of FOXO3a localization, which may contribute to the overall physiological and pathophysiological functions of this transcription factor.
FOXO3a (Foxo3 in mouse) is a member of the Forkhead box class O (FOXO) subfamily of transcription factors (Kaestner et al. 2000). Inactive FOXO3a resides in the cytoplasm, shuttling into the nucleus following dephosphorylation where it can modulate transcription of genes involved in cell differentiation, cell cycle, and metabolism (Burgering and Kops 2002; Greer and Brunet 2005). FOXO3a can also promote apoptosis after cell stress or withdrawal of growth factors via transcriptional up-regulation of pro-apoptotic proteins (Brunet et al. 1999; Gilley et al. 2003). More recently, FOXO3a activation has been associated with transcriptional inhibition, including down-regulation of nuclear-encoded genes associated with mitochondrial function (Ferber et al. 2012). In neurons, FOXO3a activation can protect against excitotoxic and proteotoxic insults (Mojsilovic-Petrovic et al. 2009) or trigger neuronal death (Gilley et al. 2003; Davila and Torres-Aleman 2008; Dick and Bading 2010). Pathologic brain activity in the form of seizures also regulates FOXO3a activation (Shinoda et al. 2004; Murphy et al. 2010).
A number of nuclear transcription factors with roles in neuronal survival have also been found within mitochondria. This includes cyclic AMP response element-binding protein (CREB) (Cammarota et al. 1999; Lee et al. 2005; Ryu et al. 2005) and NfkB (Cogswell et al. 2003). FOXO3a has been implicated in control of mitochondrial function, morphology, and biogenesis (Ferber et al. 2012), and has been reported in mitochondria in non-neuronal cells (Jacobs et al. 2008; Meshkini and Yazdanparast 2012). Here, we present evidence that FOXO3a protein is normally resident in brain mitochondria.
Materials and methods
All animal experiments were reviewed and approved by the Institutional Research Ethics Committee. Adult male rats (Sprague–Dawley, 280–320 g) and mice (C57BL/6; 20–25 g) (Harlan, Shaw Farms, Oxon, UK) were used. Briefly, animals were anesthetized and placed in a stereotaxic frame. A craniectomy was performed and a guide cannula affixed for intra-amygdala injections, as described (Schindler et al. 2006a; Murphy et al. 2007; Engel et al. 2010). Animals were then placed in an open chamber and EEG recordings from freely moving rodents initiated. Intra-amygdala injection of kainic acid (KA; 0.1 μg in 0.5 μl or 0.3 μg in 0.2 μl for rats and mice, respectively) was performed to trigger status epilepticus. Control animals received intra-amygdala vehicle phosphate-buffered saline. Seizures were recorded using a Grass digital EEG from skull-mounted recording screws (Bilaney Consultants Ltd, Kent, UK) placed above the left and right hippocampi, with a third as a reference over the occipital area (Engel et al. 2010; Murphy et al. 2010). Animals were administered lorazepam (6 mg/kg) 40 min later and then killed at various times, followed by saline perfusion. Hippocampus was removed and processed whole or microdissected to obtain enriched CA3, CA1, and dentate gyrus portions, as described (Jimenez-Mateos et al. 2012). For collection of organs, animals were killed under deep barbiturate anesthesia and saline perfused.
Human brain samples
This study was approved by the Legacy Health System Institutional Review Board, and informed consent was obtained from all patients. Age- and gender-matched control tissue was from the Brain and Tissue Bank for Developmental Disorders at the University of Maryland, Baltimore, MD. Patient and autopsy control data (n = 3 per group) have been previously reported (Schindler et al. 2006a). Samples were processed to obtain purified mitochondrial fractions as described (Schindler et al. 2006a).
Foxo3 immunogold labeling and electron microscopy was performed as previously described, with modifications (Henshall et al. 2002). Briefly, rats (n = 1 or 2 per group) were killed and transcardially perfused with 2% glutaraldehyde, 1% paraformaldehyde, 0.1% picric acid, and 0.1M HEPES. Hippocampus was embedded in LR white, mounted on nickel grids, blocked, and incubated with anti-Foxo3 antibody (rabbit polyclonal, Cell Signaling Technology, Beverly, MA, USA). Samples were then incubated with 5 nm goat anti-rabbit immunogold antibody (BB International, Cardiff, UK). Samples were examined using a Hitachi H7650 transmission electron microscope.
Procedures were as previously described, with modifications (Henshall et al. 2004; Schindler et al. 2006b). Tissue or cultured cells were disrupted in M-SHE buffer (210 mM mannitol, 70 mM sucrose, 10 mM HEPES-KOH pH 7.4, 1 mM EDTA, 1 mM EGTA) containing protease and phosphatase inhibitor cocktails (Sigma-Aldrich Ireland, Arklow, Ireland). Samples were homogenized and centrifuged to spin down nuclei and unbroken cells. Samples were then centrifuged at 10 000 g for 15 min to separate mitochondria from the cytoplasm, and the mitochondrial extract was then resuspended twice in M-SHE buffer and centrifuged at 10 000 g for 15 min, and the cytosolic fraction was centrifuged twice at 10 000 g for 15 min. Verification of the purity of cytoplasmic and mitochondrial fractions from human temporal lobe epilepsy (TLE) brain has been reported previously using standard markers (Henshall et al. 2004; Schindler et al. 2006b).
HT22 cells (a gift from C. Culmsee) were grown in Dulbecco's Modified Eagle Medium (DMEM, Lonza Biologics, UK), 4.5 g/L high glucose with glutamine and sodium pyruvate containing 10% fetal bovine serum, supplemented with penicillin/streptomycin, and incubated in 5% CO2 humidified atmosphere at 37°C. Hippocampal primary neurons were cultured from E18 mouse embryos and seeded at a density of approximately 10 000 cells per well in 24-well plates (Jimenez-Mateos et al. 2012).
Western blotting was as described (Henshall et al. 2002; Schindler et al. 2006a,b). Samples were boiled in gel-loading buffer and separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) followed by transfer of proteins to membranes and incubation with antibodies against: GAPDH (Cell Signaling Technology, Beverly, MA, USA), Porin (Merck, Darmstadt, Germany), FOXO3a and Lamin A/C (Cell Signal. Tech.), pan-14-3-3, 14-3-3θ and α-Tubulin (Santa Cruz Biotechnology, Santa Cruz, CA, USA), and COXIV (Oxphos International). Membranes were then incubated with horseradish peroxidase-conjugated secondary antibodies (Cell Signal. Tech.), followed by chemiluminescence detection and visualization.
For chromatin immunoprecipitation (ChIP) analysis (Shinoda et al. 2004; Fukuoh and Kang 2009), mitochondria from mouse hippocampus were sonicated to shear mitochondrial (mt) DNA and incubated with ChIP-grade rabbit anti-FOXO3a antibody (1 : 100, Cat #12162; Abcam, Cambridge, UK) or normal rabbit IgG, and protein A/G agarose beads (Santa Cruz Biotechnology). Protein–mtDNA complexes were collected and DNA extracted and denatured. PCR was performed using 2 μl of DNA per reaction. Primers used for the D-loop: Forward, 5′ – ggggccaaccagtagaaca – 3'; reverse, 5′ – aataccagctttgggtgctg – 3′. The primers used were for ND6: Forward, 5′ – accaatctccaaaccatca – 3′; reverse, 5′ – gggtttggtggatcgttttt – 3′.
Proteinase K protection assay
Based on reported techniques (Lee et al. 2005), mitochondrial fractions from HT22 cells in M-SHE buffer were split into aliquots and incubated with either Triton X-100 (10%), Proteinase K (0.6 mg), or both, for 30 min on ice. After incubation, 5 μL of protease inhibitor (Sigma-Aldrich) and gel-loading buffer were added and samples were boiled for western blot analysis of Foxo3.
Foxo3 plasmids and transfections
The Foxo3–EGFP plasmid encoding the mouse Foxo3 gene followed by enhanced green fluorescent protein (GFP) has been previously described (Jacobs et al. 2003). pEGFP-N1 was used as control plasmid (Clontech Laboratories Inc., Mountain View, CA, USA). A mitochondrially targeted version of the Foxo3–GFP plasmid was generated by inserting the mitochondrial targeting sequence (mts) from subunit VIII of human cytochrome c oxidase. The amplification was performed using the pDsRed2-Mito plasmid (Clontech Laboratories Inc., Mountain View, CA, USA) as a template. The primers used were as follows: forward 5′ – gaattcatgtccgtcctgacgccgct –3′ and reverse 5′ – gtcgaccccaacgaatggatcttggcgc – 3′. Both the amplified mitochondrial targeting sequence and Foxo3–GFP plasmids were digested with EcoRI and SalI Fastdigest restriction enzymes (Fermentas, Ireland). Plasmid and PCR products (5 : 1) were digested at 37°C for 5 min with both enzymes. The ligation of the PCR product in the opened Foxo3–GFP plasmid was performed using T4 DNA ligase (Fermentas, Ireland). Ligation products were purified prior to sequencing and the preparation of maxi-preps. Plasmid DNA was mixed with Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) and added to serum-free medium for 4 h at 37°C. Transfection medium was then replaced by growing medium and experiments performed 24 h later.
Cell viability assays
HT22 cells were seeded at 4500 cells per well with or without plasmids and then treated with monosodium glutamate (Sigma-Aldrich) (Tobaben et al. 2011). After incubation, 50 μl of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), 0.5% w/v in phosphate-buffered saline was added to each well for 4 h. Medium was then discarded and MTT crystals dissolved in dimethyl sulfoxide and absorbance read at 560 nm. Primary neurons were transfected after 11 days in vitro with 1.25 μg of plasmid per well using calcium phosphate technique (Jimenez-Mateos et al. 2012). Twenty-four hours after transfection, neurons were treated with glutamate/glycine (100 μM/10 μM) for 30 min. Twenty -four hours later, cell death assay was performed with live staining with Hoechst 33258 (1 μg/mL) and propidium iodide (5 μM) in the medium for 30 min.
HT22 cells were stained with Mitotracker red CMXRos for 30 min, then washed, fixed, mounted, and imaged using confocal microscopy (510 META, 63× objective, oil immersion) (Carl Zeiss, AG, Oberkochen, Germany).
Data are presented as mean ± SEM. Data were analyzed using analysis of variance (anova) with post hoc Tukey test (StatView software, SAS Institute, inc., Cary, NC, USA). Significance was accepted at p < 0.05.
Foxo3 localization in mitochondria
We previously reported the activation of Foxo3 in adult rat hippocampus after status epilepticus (Shinoda et al. 2004), but the subcellular localization of Foxo3 was not investigated. We therefore began by using transmission electron microscopy. As expected, immunolabeled Foxo3 gold beads were detected in the cytoplasm of hippocampal neurons in control rat hippocampus (data not shown). Foxo3-labeled gold beads were also present in mitochondria of neurons in control hippocampus (Fig. 1a–d). Foxo3-labeled gold beads were also detected in mitochondria in non-damaged hippocampal neurons in rats after status epilepticus (Fig. 1e, f).
To support these data, we examined fractions of rat hippocampus prepared by density centrifugation. Immunoblotting for known compartment-specific markers – 14-3-3θ for cytoplasm, Lamin A/C for nucleus, and Porin for mitochondria – confirmed purity of fractions (Fig. 2a). Foxo3 protein was detected in the mitochondrial fraction of rat hippocampal subfields (Fig. 2b) as well as in the cytoplasm and nucleus (data not shown). Using the same centrifugation technique, we prepared subcellular fractions from C57BL/6 mouse brain and a selection of other mouse organs (Fig. 2c). Immunoblotting detected Foxo3 in the mitochondrial compartment of mouse brain and several other mouse tissues tested (Fig. 2d).
Foxo3 binds mtDNA
Next, we investigated whether Foxo3 could bind mtDNA, as was reported for mitochondrial NfkB (Cogswell et al. 2003) and CREB (Lee et al. 2005). The core recognition motif for FOXO proteins is GTAAA(C/T)A (Greer and Brunet 2005), and bioinformatics analysis identified a potential consensus within the D-loop of the rat, mouse, and human mitochondrial genome (Fig. 2e). To determine if Foxo3 was capable of binding this site, we adapted a previous Foxo1 ChIP assay (Shinoda et al. 2004). An amplification product corresponding to the D-loop region containing the Foxo3 consensus site was immunoprecipitated from mouse brain mitochondria by anti-Foxo3 antibodies, which was absent from IgG controls and when the ND6 region was analyzed (Fig. 2f).
Localization of Foxo3 within mitochondrial matrix
The detection of Foxo3 binding to mtDNA suggests that the protein may reside within the mitochondrial matrix. To confirm this, subcellular fractions were prepared from HT22 cells (Fig. 3a), a well-characterized mouse hippocampal cell line (Tobaben et al. 2011). As in mouse and rat brain, Foxo3 was present in purified mitochondria from HT22 cells (Fig. 3b). Incubation of mitochondria with proteinase K alone failed to degrade mitochondrial Foxo3 (Fig. 3b, lane 2). In contrast, when mitochondrial fractions were incubated with the detergent Triton X-100 in addition to proteinase K, Foxo3 protein was degraded (Fig. 3b, lane 3).
Foxo3–GFP fusion protein partially localizes to mitochondria
We next examined Foxo3 distribution after transfecting HT22 cells with Foxo3–GFP (Jacobs et al. 2003). Transfection efficiency was in the region of ~50–60% (data not shown). In whole-cell lysates, the Foxo3–GFP fusion protein was detected at its predicted weight, ~30 kD above the endogenous Foxo3 protein in HT22 cells (Fig. 3c). Foxo3–GFP was also present within the cytoplasmic and in mitochondrial fractions of HT22 cells (Fig. 3c). Confocal microscopy supported this distribution showing Foxo3–GFP colocalization with mitotracker, whereas GFP-only transfected cells exhibited a diffuse somal pattern and no specific colocalization with the mitochondrial marker (Fig. 3d).
A mitochondrially targeted Foxo3–GFP protein protects against glutamate-induced cell death in HT22 cells
We next examined the effects of enhancing mitochondrial Foxo3 levels using a mitochondrial targeting sequence (mts) in a model of glutamate-induced neuronal apoptosis (Landshamer et al. 2008). Consistent with previous work (Tobaben et al. 2011), HT22 cells displayed dose-dependent cell death 17 h after glutamate treatment (Fig. 4a). Western blot analysis of purified mitochondrial fractions from HT22 cells confirmed increased Foxo3 levels in cells transfected with the mts–Foxo3–GFP plasmid relative to the original Foxo3–GFP plasmid (Fig. 4b).
Next, we transfected each plasmid into HT22 cells and studied the effects on glutamate-induced neuronal death. Assessment of cell death 17 h after treatment with 1 mM glutamate found no effect of either Foxo3–GFP or mts-Foxo3–GFP transfection (data not shown). At 3 mM glutamate, the mts–Foxo3–GFP, but not Foxo3–GFP, increased cell survival (p < 0.05 compared with Foxo3–GFP and GFP-only controls; data not shown). At 5 mM glutamate, Foxo3–GFP protected against glutamate-induced cell death compared with GFP-only transfected cells, while mts–Foxo3–GFP increased HT22 cell survival compared with both Foxo3–GFP and GFP-only transfected cells (Fig. 4c). This was a time-dependent effect, however, because Foxo3–GFP and mts–Foxo3–GFP did not protect against HT22 cell death assessed earlier (15 h) after 5 mM glutamate treatment (data not shown). Transfection of either plasmid into mouse primary hippocampal neurons did not protect against glutamate-induced cell death (Fig. 4d).
Reduction of mitochondrial Foxo3 levels in seizure-damaged mouse hippocampus
To expand on these in vitro findings, we examined Foxo3 responses to an in vivo excitotoxic insult. Mice underwent prolonged seizures induced by intra-amygdala KA, resulting in extensive hippocampal damage (Engel et al. 2010; Murphy et al. 2010). Whole-cell levels of Foxo3 were not changed up to 8 h after status epilepticus (Fig. 4e). In contrast, mitochondrial levels of Foxo3 were reduced at 4 and 8 h after status epilepticus (Fig. 4f).
Increased mitochondrial FOXO3a levels in human temporal lobe epilepsy
Finally, we examined FOXO3a levels in the mitochondrial fraction of human control hippocampus and hippocampus surgically obtained from patients who underwent surgery for pharmacoresistant TLE. These samples have previously been validated as mitochondrial and containing no significant cross-compartment contamination using fraction-specific markers (Henshall et al. 2004; Schindler et al. 2006a). Western blotting revealed FOXO3a in the mitochondrial compartment of control human hippocampus (Fig. 4g). FOXO3a protein was also present in the mitochondrial fraction of TLE hippocampus, and densitometry-determined levels were significantly higher (p < 0.05 compared to control, Fig. 4g).
A number of nuclear transcription factors have been found to be present within mitochondria, including CREB and NfkB (Lee et al. 2005; Ryu et al. 2005) among others. CREB and NfkB both appear to directly interact with mtDNA and regulate transcription of mtDNA-encoded genes (Cogswell et al. 2003; Lee et al. 2005; Ryu et al. 2005). Experiments here using tissue and cells from mice as well as rat and human hippocampus suggest that FOXO3a joins this group of transcription factors. FOXO3a appears to be normally resident in neuronal mitochondria, as evidenced by electron and confocal microscopy, subcellular fractionation analysis, and a proteinase K protection assay. As reported for CREB (Lee et al. 2005; Ryu et al. 2005) and NfκB (Cogswell et al. 2003), there is a putative binding site for FOXO3a in the D-loop region of the mitochondrial genome, and we detected binding of FOXO3a to mtDNA. While this is the first study to show a mitochondrial localization of FOXO3a in brain, FOXO3a protein has been reported in mitochondrial fractions from non-neuronal cells (Jacobs et al. 2008; Meshkini and Yazdanparast 2012). Previous immunostaining of FOXO3a in non-neuronal cells also featured distribution consistent with mitochondrial localization (Dehner et al. 2008; Srivastava et al. 2010). FOXO3a was not among the proteins detected by proteomic analysis of mitochondria (Mootha et al. 2003; Taylor et al. 2003). However, neither CREB nor NfkB components, known to be present in mitochondria (Cogswell et al. 2003; Lee et al. 2005; Ryu et al. 2005), were detected in those studies. Thus, converging evidence supports a mitochondrial localization of FOXO3a in brain.
Prolonged glutamate treatment of HT22 cells triggers apoptosis as a result of glutathione depletion and reactive oxygen species (Tobaben et al. 2011). In this study, we found that Foxo3 over-expression conferred a small protective effect against glutamate toxicity in HT22 cells, although increased protein level does not necessarily equate to increased activity. We also found that enhancing mitochondrial loading of Foxo3 promoted the protective effects of Foxo3 against high-dose glutamate-induced injury in these cells. The protective effect was observed, however, at only one dose in HT22 cells, and no effect was found in primary hippocampal neurons from mice. Thus, increasing mitochondrial Foxo3 levels in certain cells and under a restricted set of conditions can have small protective effects against glutamate toxicity. Some of our data are therefore consistent with reports of protective effects of FOXO3a against excitotoxic injury in neurons (Mojsilovic-Petrovic et al. 2009). Nevertheless, Foxo3 or enhanced mitochondrial Foxo3 do not appear to be broadly protective against glutamate-induced neuronal death in mouse cells. Loss of mitochondrial FOXO3a or inhibition of FOXO3a upload into mitochondria may, however, have a greater effect on cell viability. This could be tested by interfering with the process by which FOXO3a enters or leaves mitochondria.
Notably, this study shows that mitochondrial FOXO3a levels are affected by pathologic brain activity. The reduction in mitochondrial Foxo3 protein after status epilepticus in mice is most likely explained by translocation out of mitochondria as overall cell levels of Foxo3 did not change. The apparent release of Foxo3 from mitochondria is reminiscent of the release of apoptogenic proteins such as apoptosis-inducing factor (AIF) and cytochrome c, which occur in this model (Murphy et al. 2007; Engel et al. 2010). If released Foxo3 is transcriptionally active, this may contribute to the pool that regulates expression of nuclear-encoded genes. It is also notable that IκB, which along with NfκB is normally resident in mitochondria, is released from mitochondria during apoptosis (Cogswell et al. 2003) and lower mitochondrial CREB may be associated with neurodegeneration (Lee et al. 2005). In contrast to the mouse data, FOXO3a levels were higher in mitochondria in patients with TLE. Thus, mitochondrial FOXO3a levels are regulated in opposite directions according to whether seizures are prolonged and harmful or brief and repetitive, consistent with a model where higher mitochondrial FOXO3a levels may be protective. Activation of FOXO3a has been associated with increased neuronal excitability (Howlett et al. 2008), so the increased FOXO3a in mitochondria may also influence excitability. Taken together, our data identify a novel localization of FOXO3a and potentially support a role for mitochondrial FOXO3a in the molecular neuropathology of TLE.
The authors thank Caoimhin Concannon, Heiko Duessmann, Monika Jarzabek, Jochen Prehn, Lavanya Ramapathiran, and Petronela Wiesova for suggestions and technical advice. The authors also thank the UMBTB for the human tissue, Carsten Culmsee for the HT-22 cells, and Martin P. Smidt for the Foxo3–EGFP plasmid. This work was supported by Science Foundation Ireland grant 08/RFP/1745 (to D.C.H.) and a fellowship to T.E. from the Health Research Board (HRB PD/2009/31). The authors have no conflict of interest to declare.