A major phenotype seen in neurodegenerative disorders is the selective loss of neurons due to apoptotic death and evidence suggests that inappropriate re-activation of cell cycle proteins in post-mitotic neurons may be responsible. To investigate whether reactivation of the G1 cell cycle proteins and S phase entry was linked with apoptosis, we examined homocysteine-induced neuronal cell death in a rat cortical neuron tissue culture system. Hyperhomocysteinaemia is a physiological risk factor for a variety of neurodegenerative diseases, including Alzheimer’s disease. We found that in response to homocysteine treatment, cyclin D1, and cyclin-dependent kinases 4 and 2 translocated to the nucleus, and p27 levels decreased. Both cyclin-dependent kinases 4 and 2 regained catalytic activity, the G1 gatekeeper retinoblastoma protein was phosphorylated and DNA synthesis was detected, suggesting transit into S phase. Double-labelling immunofluorescence showed a 95% co-localization of anti-bromodeoxyuridine labelling with apoptotic markers, demonstrating that those cells that entered S phase eventually died. Neurons could be protected from homocysteine-induced death by methods that inhibited G1 phase progression, including down-regulation of cyclin D1 expression, inhibition of cyclin-dependent kinases 4 or 2 activity by small molecule inhibitors, or use of the c-Abl kinase inhibitor, Gleevec™, which blocked cyclin D and cyclin-dependent kinase 4 nuclear translocation. However, blocking cell cycle progression post G1, using DNA replication inhibitors, did not prevent apoptosis, suggesting that death was not preventable post the G1-S phase checkpoint. While homocysteine treatment caused DNA damage and activated the DNA damage response, its mechanism of action was distinct from that of more traditional DNA damaging agents, such as camptothecin, as it was p53-independent. Likewise, inhibition of the DNA damage sensors, ataxia-telangiectasia mutant and ataxia telangiectasia and Rad3 related proteins, did not rescue apoptosis and in fact exacerbated death, suggesting that the DNA damage response might normally function neuroprotectively to block S phase-dependent apoptosis induction. As cell cycle events appear to be maintained in vivo in affected neurons for weeks to years before apoptosis is observed, activation of the DNA damage response might be able to hold cell cycle-induced death in check.
Apoptosis or programmed cell death plays a critical role in the development of the mammalian nervous system, but in mature neurons apoptotic mechanisms are held in check. In many neurodegenerative disorders a major phenotype is the selective apoptosis of specific subsets of neuronal populations over years or even decades (Busciglio and Yankner, 1995; Offen et al., 2000; Jimenez del Rio and Velez-Pardo, 2001; Raina et al., 2001; Becker and Bonni, 2004; Jellinger, 2006; Zhu et al., 2006). Alzheimer’s disease affects up to 15% of people over the age of 65 and is characterized by the triad of amyloid plaques, neurofibrillary tangles and neuronal cell death. Memory loss and cognitive impairment are caused by neurodegeneration in the hippocampus and cerebral cortex in patients with Alzheimer’s disease. The cause of this neurodegeneration is under active debate and molecules implicated in the pathogenesis of Alzheimer’s disease include beta-amyloid protein, tau, presenilins, homocysteine, apolipoprotein E and most recently, the proteins that regulate cell cycle progression.
The cell cycle is governed by the action of the cyclin-dependent kinase complexes (cyclin–cdks) (Malumbres and Barbacid, 2009). Cyclins bind to cdks and positively regulate their kinase activity (Sherr and Roberts, 1999). In mammalian cells, cyclin D-cdk4/6 and cyclin E–cdk2 regulate the G1–S phase transition. These kinases phosphorylate substrates, such as the tumour suppressor retinoblastoma protein (Rb), which acts as a brake to S phase entry. In its unphosphorylated state, Rb sequesters transcription factors required for S phase entry, and cells are arrested in G1. Because cyclin–cdks are responsible for all cell cycle transitions, cdk activity is tightly regulated by a combination of mechanisms, including changes in the cyclin level, nuclear to cytoplasmic shuttling, phosphorylation of positive and negative regulatory sites on the cdk partner and interaction with stoichiometric inhibitors (cyclin-dependent kinase inhibitors), such as p27Kip1, p21Cip1, p15Ink4b or p16Ink4a (Sherr and Roberts, 1999). Interaction of cyclin-dependent kinase inhibitors with cdks is the major route by which antiproliferative responses, such as contact, tumour growth factor-β, ultraviolet irradiation or differentiation, affect cdk activity (Besson et al., 2008).
The differentiated neurons in the CNS had been described as post-mitotic cells, believed to have lost their capacity for proliferation. However, loss of cell cycle inhibitors, such as p27 or p19 in knockout animals, caused immature CNS neurons to re-enter the cell cycle, demonstrating that these neurons were being actively maintained by cyclin-dependent kinase inhibitors in a reversible, quiescent state (Zindy et al., 1999; Akashiba et al., 2006a, b; Ueyama et al., 2007). Increasing evidence suggests a correlation between cell cycle re-entry and the neuronal apoptosis seen in neurodegenerative diseases. Evidence for cell cycle reactivation is found both in Alzheimer’s disease mouse models and Alzheimer’s disease brains (Copani et al., 1999; Chen et al., 2000; Liu and Greene, 2001; Becker and Bonni, 2004). Altered and inappropriate cyclin D, cdk4 and p27 overexpression and Rb phosphorylation was detected in Alzheimer’s disease neurons as compared to age-matched controls (McShea et al., 1997; Nagy et al., 1997, 1998; Raina et al., 2000; Ogawa et al., 2003; Thakur et al., 2008). DNA replication was detected by fluorescent in situ hybridization in neurons from Alzheimer’s disease model mice (Angelastro et al., 2001; Yang et al., 2003), but was rarely observed in either normal aged brain tissue or in the naturally occurring apoptosis that occurs during development (Herrup and Yang, 2007). Activated DNA replication proteins, such as the mini-chromosome maintenance (MCM2) proteins, have been detected in Alzheimer’s disease brains, indicating that inappropriate S phase entry might occur during disease progression (Bonda et al., 2009). In tissue culture models, cell cycle re-entry or cyclin–cdk activation was found associated with the apoptosis seen in PC12 cells and sympathetic neurons in response to deprivation of trophic support (Park et al., 1996), exposure to beta-amyloid protein (Giovanni et al., 1999) or camptothecin treatment (Park et al., 1997). Apoptotic stimulation of cortical neurons caused cyclin D1 to re-locate to the nucleus (Sumrejkanchanakij et al., 2003), and p27 levels were reduced in several neuronal death models including cerebral ischaemia, repolarization and glutamate toxicity (Akashiba et al., 2006a, b; Qiu et al., 2009). The cell cycle reactivation model suggests that instead of completing the entire cell cycle, the passage into S phase induces apoptosis, contributing to the pathology of diseases such as Alzheimer’s disease. However, it is unclear whether cell cycle reactivation is the initiating event that causes neuronal death, or merely a neuroprotective response to apoptosis induction (Raina et al., 2004; Herrup and Yang, 2007).
In this study, we examined the G1 phase of the cell cycle, to determine whether the exit from G0 phase and transit into S phase were responsible for homocysteine-induced neuronal apoptosis in cortical neurons. Homocysteine is a non-essential sulphur-containing amino acid, produced as a natural consequence of methionine metabolism and has been shown to be a strong independent risk factor for dementia and other neurodegenerative diseases (Clarke, 1998; Seshadri et al., 2002; Mizrahi et al., 2004; Agnati et al., 2005). Our results demonstrate that the G1 cell cycle proteins are rapidly activated following homocysteine treatment and neurons could be protected from homocysteine-induced cell death by a variety of methods that inhibit G1 phase progression. We found a 95% co-localization of apoptosis markers with DNA synthesis markers, demonstrating that those cells that entered S phase underwent apoptosis. Interestingly, inhibition of the DNA damage response [ataxia-telangiectasia mutant (ATM) and ataxia telangiectasia and Rad3-related protein (ATR)] did not rescue homocysteine-induced apoptotic death, but instead exacerbated it. Thus, our data suggest that G1 transit and S-phase entry are critical factors in homocysteine-induced neuronal apoptosis, and the ATM/ATR response may normally function neuroprotectively to prevent or repair inappropriate cell cycle progression.
Materials and methods
Homocysteine, caffeine, camptothecin, olomoucine, 5-bromo-2-deoxyuridine (BrdU), aphidicolin, hydroxyurea, nocodazole and poly-l-lysine were purchased from Sigma (St. Louis, MO). Ku-55933, K2 inhibitor II (Davis et al., 2001) and K4 inhibitor II (Kubo et al., 1999) were purchased from Calbiochem (Gibbstown, NJ). Gleevec™ (STI571 mesylate salt) was a gift from Novartis Pharmaceuticals. Camptothecin, olomoucine, Ku-55933, K2 inhibitor II and K4 Inhibitor II were dissolved in dimethyl sulphoxide. Caffeine was dissolved freshly in medium. Control cultures contained <0.3% dimethyl sulphoxide as a vehicle control, and were not affected within the time range of the experiments described. Homocysteine was prepared at 100 mM in fresh medium. Ku-55933, olomoucine, K2 inhibitor II and K4 inhibitor II were diluted with medium and added into culture 1 h before homocysteine addition.
Monoclonal anti-microtubule associated protein 2, actin and anti-BrdU antibodies were purchased from Sigma (St. Louis, MO). Phospho-checkpoint protein kinase (Chk) 1 (Ser296), phospho-Chk1 (S319), phospho-Rb (Ser795), phospho-p53 (S15), phospho-cdk2 (160), phospho-ATR (S428) and activated caspase-3 (Asp175) antibodies were purchased from Cell Signalling (Danvers, MA). Antibodies to c-Abl (K-12), cyclin D1 (H-295), cdk4 (C-22) and cdk2 (sc-163) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Phospho-histone H2A.X (S139) (clone JBW301) antibody was purchased from Millipore (Billerica, MA). Antibodies to cyclin A2 and phospho Chk2 (T68) were purchased from AbCam (Cambridge, MA). p27Kip1 antibodies were purchased from BD transduction laboratories (San Jose, CA). Fluorescent-conjugated Alexa Fluor 488, 568 and 632 anti-mouse and anti-rabbit immunoglobulin G were purchased from Molecular Probes (Carlsbad, CA).
Primary culture of rat cerebral cortical neurons
Time-mated pregnant Sprague Dawley rats were obtained from Charles River Laboratories (Wilmington, MA). Primary neuronal cultures of rat cerebral cortex were prepared as described with some modifications (Carey et al., 2002). In brief, whole brains were isolated from foetal rats at embryonic days 17–18 and the cerebral cortices were treated with 0.25% trypsin (Cellgro, Manassas, VA) and 0.01% deoxyribonuclease I (Sigma) at 37°C for 15 min. Cells were dissected, mechanically dissociated and suspended in Neurobasal medium (Invitrogen, Carlsbad, CA) containing 10% foetal bovine serum. Cortical cells were plated at a density of 1.2 × 107 cells in 100 mm dishes, 4 × 105 cells in 12-well dishes, 105 cells in 48-well plates or on glass coverslips precoated with poly-l-lysine. Cultures were maintained at 37°C in humidified 5% CO2, 95% air. The day after plating, the medium was changed to serum-free Neurobasal medium supplemented with 2% B27 (Invitrogen), 0.5 mM glutamine and 0.002% gentamycin sulphate. Fifty percent of the culture medium was changed every 3 days. Homocysteine or other reagents were added into medium at Days 7–9, as indicated in the text.
Neuronal survival was determined using the CellTiter-Glo Luminescent Cell Viability Assay from Promega (Madison, WI) per the manufacturer’s instructions. Briefly, cells were maintained at room temperature for 10 min, mixed with 40 µl/well CellTiter-Glo reagent for 5 min and incubated for an additional 10 min at room temperature. The luminescence of the lysate was measured with a Glomax luminometer (Promega, Madison, WI). Luminescent intensity was proportional to the number of metabolically viable neurons (Crouch et al., 1993; Andreotti et al., 1995). Each experiment was performed at least three independent times. Neuronal survival after treatment with pharmacological agents was compared to parallel treatment with control medium, and survival was expressed as percent of control ± standard deviation (SD) (n = 6 per condition).
Immunoblot and immunoprecipitation analysis
Assays were performed as previously described (James et al., 2008). The preparation of a cytoplasmic and nuclear extraction was done using the CellLytic Nuclear Extraction Kit (Sigma, St. Louis, MO) according to manufacturer’s instructions.
Cells cultured on poly-l-lysine-coated glass coverslips or multi-well plates were treated as indicated, fixed with 4% formaldehyde (12 min at room temperature), permeabilized in phosphate buffered saline-0.25% Triton X-100 (5 min at room temperature) and stained with the first antibodies, as specified in the figure legends. They were then visualized by incubation with Alexa Fluor 488, 568, 632 goat anti-mouse IgG or goat anti-rabbit IgG, followed by DNA staining with either ToPro 3 (Molecular Probes), Hoechst 33342-bisbenzimide (Sigma), DapI (Sigma) or propidium iodide (Sigma). For anti-cyclin D1, anti-cdk4 or anti-phospho-Rb staining, cells were stained with the primary antibodies at 4°C overnight, followed by peroxidase-conjugated secondary antibodies (SouthernBiotech, Birmingham, Al). Immunostains were visualized using a fluorescent tyramide reagent (TSA-direct NEL-701, PerkinElmer, Waltham, MA) as described previously (Ye et al., 2009). Confocal images were acquired by a Zeiss 510 laser-scanning microscope.
Comet assay for DNA damage
The CometAssay Kit (Trevigen, Gaithersburg, MD) was used to detect single- and double-stranded DNA breaks in cultured cortical neurons (Tice et al., 2000). Slides were viewed (excitation 425–500 nm) with a Zeiss Axiophot epifluorescence microscope. Two CometSlides were used for each condition. In normal neurons, the fluorescence is confined mostly to the nucleus because undamaged DNA cannot migrate. In neurons with DNA damage, DNA is denatured by the alkali solution used for single-strand break detection, or the neutral solution used for double-strand break detection, and the negatively charged DNA fragments are released from the nucleus and migrate toward the anode.
Terminal deoxynucleotidyl transferase-mediated dUTP-biotin nick end labelling assay
For dual staining with the terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labelling (TUNEL) reagents and BrdU, or dual staining with TUNEL and anti-activated caspase-3 labelling, cells were cultured on glass coverslips, fixed in 4% paraformaldehyde (15 min at room temperature) and permeabilized by incubation with phosphate buffered saline–0.25% Triton X-100 (5 min). Cells were stained with the TUNEL reagents (DeadEnd Fluorometric TUNEL System, Promega) first, as previously described (Ye and Zhang, 2004). For the BrdU staining, after the last wash cells were incubated with 2 M HCl at room temperature for 30 min to denature DNA, washed with phosphate buffered saline, incubated with mouse anti-BrdU (1:75 dilution, diluted in 1% bovine serum albumin/phosphate buffered saline) antibodies (Sigma, St. Louis, MO) for 1 h, washed with phosphate buffered saline and visualized by incubation with Alexa Fluor 593 rabbit anti-mouse IgG (1:2000 diluted in 1% bovine serum albumin/phosphate buffered saline) for 1 h at room temperature. For anti-activated caspase-3 immunoflourescent labelling, cells were incubated with mouse anti-activated caspase-3 antibody, visualized by incubation with Alexa Fluor 593 rabbit anti-mouse IgG and washed with phosphate buffered saline, before slides were mounted with ProLong Gold antifade reagent (Molecular Probes).
Cyclin D1 knockdown
Cyclin D1 antisense oliogodeoxynucleotides 5′-GGAGCT GGT GTT CCA TGG-3′ were complementary to the translation start site of the cyclin D1 complementary DNA and sense oligomers 5′-CCA TGG AAC ACC AGCTCC-3′ were used as a control (Shuai et al., 2006). The two sequences were synthesized and optimized by Oligos Etc Inc. (Wolsonville, OR). Antisense oligonucleotide treatment of neurons was performed as previously described (Shan et al., 2003). Cultures were transfected for 48 h. Cyclin D1 expression was determined by immunofluorescence microscopy and immunoblot analysis. Homocysteine (0.25 mM) was added after 48 h. The TUNEL assay was used 3 days later to determine the percentage of apoptotic cells.
P-values were determined by a Student’s t-test, error bars depict SD.
We determined the effect of homocysteine treatment on cell survival in primary, differentiated cortical neurons. Cortical neurons isolated from embryonic day 18 embryos were placed in culture in Neurobasal medium for 7–10 days. On day 7, more than 90% of the cultured cells were positive for microtubule associated protein 2 staining (Fig. 1C), which is a marker specific for differentiated neurons (Mehra and Hendrickson, 1993; Martinez et al., 1997). Seven-day-old cultures were then treated with homocysteine at different doses and for the indicated times. The addition of homocysteine significantly reduced neuronal survival, in a dose- and time-dependent manner (Fig. 1A and B). Approximately 30% of the neurons died within 48 h in 0.25 mM homocysteine, and 80% died after 5 days of treatment. The 0.25 mM concentration of homocysteine was in the previously published range (Kruman et al., 2000). As cell cycle activity is often density-dependent, we examined the response to homocysteine at different densities of cultured neurons, and reduced survival was seen in all conditions (Fig. 1A). Apoptosis activity in the homocysteine-treated neurons was detected by immunolabelling and confocal microscopy using anti-cleaved caspase-3 antibodies, suggesting that the neuronal death caused by homocysteine treatment was due to apoptotic activation (Fig. 1C and D). Less than 4% of the untreated cells were positive for anti-cleaved caspase-3 labelling and this low level presumably results from the low level of spontaneous death seen in our culture conditions. Homocysteine treatment increased cleaved caspase-3 detection to ∼15% (Fig. 1C and D), which corresponded with the neuronal cell death seen in these populations (Fig. 1A and B).
To determine whether homocysteine induces cell cycle re-activation in differentiated cortical neurons, we used BrdU to label newly synthesized DNA during the S phase of the cell cycle (Gonchoroff et al., 1986). Few untreated neurons labelled with anti-BrdU antibodies, suggesting that the majority of the differentiated cortical neurons had exited the cell cycle and were not proliferating (Fig. 1C and D). A significant increase (>12%) of BrdU positive nuclei was found in the homocysteine-treated neurons (Fig. 1C and D), suggesting that these cells were actively synthesizing DNA. While in theory, DNA synthesis in response to DNA damage might increase the BrdU signal in the homocysteine treated cells, previous studies have shown that DNA repair does not result in significant BrdU labelling (Cooper-Kuhn and Kuhn, 2002), suggesting that we were detecting DNA replication. Rb regulates cell cycle progression through the G1–S phase checkpoint (Sherr and Roberts, 1995, 1999), and phosphorylation of Rb by cdks causes Rb inactivation and subsequent cell cycle progression (Knudsen and Wang, 1997). We used anti-phospho-Rb (S795) immunolabelling to measure Rb phosphorylation in the presence or absence (control) of homocysteine treatment (Fig. 1C and D). Homocysteine treatment significantly increased the proportion of anti-phospho Rb (S795) positive nuclei from 18% in untreated cells to >40% in treated cells (Fig. 1C and D), suggesting that homocysteine treatment caused cell cycle reactivation and entry into S phase.
To correlate homocysteine-induced apoptosis with cell cycle phase, we stained cells with multiple antibodies simultaneously followed by confocal microscopy immunofluorescence. Co-localization of activated caspase-3 staining with BrdU labelling was seen in the nuclei of homocysteine-treated neurons (Fig. 2A). Cells were stained with anti-BrdU (blue), anti-phospho Rb (S795) (red) and TUNEL (green), and colocalization analysis of this triple labelling was performed in the homocysteine-treated neurons (Fig. 2B). The results show that colocalization between BrdU positive and TUNEL positive nuclei was almost 95% (Fig. 2B, right panel and 2C), suggesting that cells that initiated DNA synthesis and entered S phase were also undergoing apoptosis. Within the same group, less than 20% of cells contained both TUNEL and phospho-Rb (S795) staining. Likewise, <20% of the homocysteine-treated cells colocalized the BrdU and phospho-Rb (S795) markers (Fig. 2B, left, middle panels and 2C). While ∼40% of the homocysteine-treated cells became Rb positive (Fig. 1C and D), the colocalization studies (Fig. 2B and C) suggest that Rb phosphorylation at the S795 site itself was not sufficient to induce apoptosis; and entry into S phase, as measured by BrdU incorporation and DNA synthesis, was required to cause neuronal cell death in differentiated cortical neurons. Homocysteine induced cell cycle re-entry in ∼40% of the treated cells, but a smaller percentage (15–20%) entered S phase. This S phase entry, however, seemed sufficient to initiate apoptosis.
To analyse cell cycle re-entry in the presence of homocysteine further, we examined many of the major cell cycle players that regulate the G1–S phase transition: cyclin D–cdk4, cyclin A–cdk2 and p27Kip1 (Sherr and Roberts, 1999). Differentiated neurons were treated with homocysteine and analysed at different time points post treatment (Figs 3 and 4) to determine the expression levels, subcellular locations, interaction status and kinase activity of cdks and cyclin-dependent kinase inhibitors by immunoprecipitation, immunoblot and confocal microscopy. For microscopic analysis, co-staining with 4′,6-diamidino-2-phenylindole or ToPro3 was used to define nuclei (Fig. 3). We found that cyclin D1 immunoreactivity was detected mainly in the cytoplasm in differentiated cortical neurons (Fig. 3A and C). In homocysteine-treated neurons, increased nuclear and decreased cytoplasmic cyclin D1 was detected by 2 h, which continued to increase up to 20 h, suggesting that homocysteine treatment caused a cytoplasmic to nuclear translocation, similar to the mitogen-induced cytoplasmic to nuclear translocation of cyclin D1 observed in non-neuronal lineages (Gladden and Diehl, 2005). Similar homocysteine-dependent translocations were seen with cdk4 and cdk2 (Fig. 3A). Cdk4 was localized in the cytoplasm of untreated neurons (0 h), but quickly accumulated in the nucleus in the presence of homocysteine (2 h), and was predominantly nuclear by 20 h of treatment (Fig. 3A). Cyclin A localization was nuclear in the untreated neurons, and was unaltered by homocysteine treatment. Cdk2 was predominantly cytoplasmic in untreated cells (0 h), but by 2 h of homocysteine treatment it was detected in the nucleus. Thus, in response to homocysteine treatment, cyclin D, cdk4 and cdk2 appeared to translocate to the nucleus, where they could potentially partner to form cyclin D–cdk4 and cyclin A–cdk2 complexes and phosphorylate their nuclear targets (Sherr and McCormick, 2002). Rb phosphorylation itself was detected within 2 h of homocysteine treatment, suggesting that these neurons had re-entered the cell cycle and reached the G1–S phase border (Fig. 3A).
p27Kip1 interacts with both cdk4 and cdk2-associated complexes, and is a potent inhibitor of these kinases in growth-arrested cells (Blain, 2008; James et al., 2008). It has been suggested that its high expression in differentiated cells may help to maintain cdk4 and cdk2 in their catalytically inactive forms, holding neuronal cells in the G0 quiescent state (Zindy et al., 1999). To determine the localization and expression of p27, we stained homocysteine-treated neurons with anti-p27 antibodies. p27 was detected predominately in the nucleus of untreated differentiated neurons, with weaker but detectable cytoplasmic expression observed (Fig. 3A and B). In the presence of homocysteine by 20 h total p27 staining was reduced significantly (Fig. 3A). In apoptotic cells, the down-regulation of p27 appeared to occur significantly in the nucleus (Fig. 3B). As epithelial cells and lymphocytes exit G0 phase, p27 levels decrease to lower but detectable levels, restoring catalytic activity to inactive cdk complexes (Nickeleit et al., 2007; Besson et al., 2008), and these data in differentiated neurons were consistent with the idea that these cells had re-entered the cell cycle in response to homocysteine treatment.
Homocysteine-treated cells were analysed by immunofluorescence with multiple antibodies to determine co-localization of the different G1 cell cycle proteins (Fig. 3C). Double-staining with cyclin D1 and p27 antibodies demonstrated that cyclin D1 and p27 resided in different compartments in untreated cells: cyclin D1 was predominantly in the cytoplasm, while p27Kip1 was predominately in the nucleus (Fig. 3C). After homocysteine treatment, however, both cyclin D1 and p27 could be detected in the nuclear compartment in cells with intact nuclei as detected by ToPro nuclear staining (Fig. 3C). This nuclear to cytoplasmic translocation appeared to be a phenomenon specific for cyclin D1 and not cyclin D2 (Fig. 3D). Cyclin D2 was predominantly nuclear in untreated cells, and homocysteine treatment did not alter its localization (Fig. 3D).
While p27 levels appeared to remain high and nuclear in homocysteine-treated neurons that were not undergoing apoptosis, as measured by chromatin condensation (Fig. 3C), we predicted that p27 levels might decrease specifically in those neurons that were initiating apoptosis. Using anti-p27 antibodies, we found two types of p27 staining cells: homocysteine-treated cells that appeared to retain nuclear p27 (Fig. 3D, green) or alternatively those that had significantly reduced p27 expression (Fig. 3D, white dashed circles). Lack of nuclear p27 expression appeared to correlate with apoptosis, as measured visually by chromatin condensation seen in the ToPro stained cells (Fig. 3D, white dashed circles, merge). Homocysteine-treated cells with detectable p27 staining had intact nuclei, while cells with reduced or absent p27 staining were undergoing apoptosis. This suggested that p27 loss might be a prerequisite for apoptosis in differentiated neurons.
When homocysteine-treated and untreated neurons were stained with both anti-p27 and anti-phospho-Rb antibodies, an inverse correlation was observed (Fig. 3E). Little phospho-Rb was detected in untreated cells. By 7h post treatment phospho-Rb staining was seen, but only in cells that had severely reduced p27 expression (Fig. 3E, green arrows). In cells that retained high nuclear p27 expression, Rb phosphorylation was not detected, suggesting that reduction of p27 expression correlated with the appearance of phospho-Rb as a marker of the G1–S phase transition.
To demonstrate directly that neurons with activated G1 cell cycle programs were entering S phase, treated and untreated cells were co-stained with anti-cyclin D1 and BrdU antibodies (Fig. 3F). Little BrdU staining and predominantly cytoplasmic cyclin D1 were detected in untreated cells (0 h). Homocysteine treatment, however, caused the mobilization of cyclin D1. By 7 h of treatment, cyclin D1 was predominantly nuclear and BrdU specifically co-localized in these cells. Together these data suggest that homocysteine treatment caused the translocalization of cyclin D1, cdk4 and cdk2 to the nucleus, where they could potentially interact with their nuclear targets. Concomitant with this, a reduction to lower p27 levels was seen and Rb became phosphorylated. These data suggest that the differentiated neurons had exited the G0 phase, passing through G1 phase on their way to S phase. Co-localization of these events occurred in cells that underwent apoptosis, suggesting that the reactivation of the G1 cell cycle players and their activity was required for homocysteine-induced apoptosis.
Cyclin–cdks primarily phosphorylate nuclear targets during cell cycle progression. To examine the localization of the G1 cyclin–cdk complexes further, homocysteine-treated and untreated neurons were harvested at different times post homocysteine addition, and nuclear and cytoplasmic extracts were analysed by immunoblot analysis with antibodies to cyclin A, cyclin D1, cdk4 and cdk2 (Fig. 4A). Cyclin D1 was localized predominantly in cytoplasmic extracts in untreated neurons. Homocysteine treatment decreased its total levels, while significantly increasing its detection in the nucleus (Fig. 4A, lane 1). Cdk4 and cdk2 were also detected primarily in cytoplasmic extracts in untreated neurons, but by 4 and 9 h of treatment increased nuclear detection was observed (Fig. 4A, lanes 2 and 5). Cyclin A was detected in nuclear extracts in untreated neurons and the addition of homocysteine did not alter its distribution (Fig. 4A, lane 4), consistent with results seen by immunofluorescence microscopy. Cdk2 must be phosphorylated on a conserved threonine residue (T160) in order to be catalytically active (Sethi et al., 2005). Using antibodies specific for cdk2 T160 phosphorylation, we did not detect active cdk2 in untreated cells, but it was detected in homocysteine-induced nuclear extracts within 4h of treatment, indicating that cdk2 had regained activity (Fig. 4A, lane 6), consistent with late G1 phase. Phosphorylated cdk2 was also detected in cytoplasmic extracts at 9 h of treatment, consistent with the increased activation and potential nuclear to cytoplasmic shuttling of this complex (Fig. 4A, lane 6). While the total level of Rb protein was unchanged by homocysteine treatment, nuclear (Fig. 4A, lane 8) phosphorylated Rb detection increased by 4 h of treatment (Fig. 4A, lane 7). These data were consistent with the results seen by confocal microscopy: the cyclins and cdks appeared to increase their nuclear expression and activation in the presence of homocysteine. We did not detect any cleaved caspase-3 expression in nuclear or cytoplasmic extracts in untreated neurons, but significant caspase-3 was detected in cytoplasmic extracts by 4 h of homocysteine treatment (Fig. 4A, lane 9), consistent with homocysteine’s ability to induce apoptosis following cell cycle re-entry.
Immunoprecipitation of treated and untreated lysates with cdk4 antibodies, followed by cyclin D1 immunoblot analysis, demonstrated that cyclin D1–cdk4 complex formation increased in the presence of homocysteine (Fig. 4C, lane 1). As the cdk4 monomer has minimal activity, increased complex formation should correspond to increased kinase activity (Bockstaele et al., 2006). Immunoprecipitation with cdk4 antibodies, followed by the addition of recombinant Rb substrate and γ-ATP in an in vitro kinase assay, demonstrated that homocysteine treatment increased cdk4 catalytic activity as well (Fig. 4C, lane 2).
Immunoblot analysis with antibodies specific for phosphorylated T160 cdk2 suggested that homocysteine-treatment increased cdk2 catalytic activity as well (Fig. 4A, lane 6). To demonstrate this directly, immunoprecipitation with cdk2 antibodies, followed by the addition of either recombinant Rb or Histone H1 substrates and γ-ATP in in vitro kinase assays, demonstrated that cdk2 became catalytically active following homocysteine-treatment (Fig. 4C, lanes 4 and 5). Immunoblot analysis of total p27 levels demonstrated that p27 expression, which was high in untreated cells, decreased following homocysteine treatment (Fig. 4B), and this loss of p27 corresponded to a concomitant reduction of p27 in cdk2-associated complexes, as detected by cdk2 immunoprecipitation (Fig. 4C, lane 3). In untreated neurons, significant p27 was associated with cdk2, but this decreased to undetectable levels by 8 h of homocysteine treatment (Fig. 4C, lane 3). As p27 is a constitutive cdk2 inhibitor (Besson et al., 2008), the lack of p27 in the cdk2 complex permits reactivation of catalytic activity, causing Rb phosphorylation and S phase progression. Our data suggest that homocysteine treatment rapidly reactivates cdk4 and cdk2 activity, permitting G1–S phase transitioning and Rb phosphorylation.
We postulated that if G1 cdk activation was required for apoptosis, inhibition of cdk activity might prevent homocysteine-induced neuronal death. We pre-treated neurons with several small molecule cdk inhibitors before the addition of homocysteine (Fig. 5A). Olomoucine is an inhibitor of cdks 1/2/5 and extracellular-signal-regulated kinase 1/microtubule associated protein kinase (Vesely et al., 1994). It is a poor inhibitor of other protein kinases, and has been shown to prevent death due to nerve growth factor deprivation of differentiated PC12 cells and sympathetic neurons (Park et al., 1996). K2 inhibitor II (Davis, et al., 2001) and K4 inhibitor II (Kubo et al., 1999) selectively inhibit cdk2 and cdk4, respectively. Pretreatment with olomoucine, K2 inhibitor II or K4 inhibitor II all protected neurons from homocysteine-induced death (Fig. 5A). Only 70% of cells remained viable after 48 h in the presence of 0.25 mM homocysteine. Treatment of cells with cdk inhibitors, however, increased cell survival (96.3% + K2 inhibitor II, P < 0.002; 93% + K4 inhibitor II, P = 0.01; or 92% + olomoucine, P < 0.005) (Fig. 5A, left), suggesting that cdk activity was required for homocysteine-dependent apoptosis. K2 inhibitor II also rendered cells resistant to the neuronal death induced at higher concentrations of homocysteine (96.5% cells survival + K2 inhibitor II compared with 56.2% cells survival in 1 mM homocysteine alone) (Fig. 5A, right).
Immunoprecipitation with cdk4 antibodies, followed by cyclin D1 immunoblot analysis demonstrated that the homocysteine-dependent increase in cyclin D–cdk4 complex formation persisted in the presence of K2 Inhibitor II, when cdk2 was inhibited (Fig. 4C, lane 1), although catalytic activity of this complex was reduced (Fig. 4C, lane 2). Co-treatment with homocysteine and the K2 inhibitor II, however, prevented the homocysteine-dependent decrease in p27 levels (Fig. 4B), and immunoprecipitation of lysates with cdk2 antibodies, followed by p27 immunoblot analysis, demonstrated that p27 remained associated with cdk2 complexes (Fig. 4C, lane 3). p27 itself is a substrate of cdk2, and its phosphorylation by cdk2 increases its proteasomal degradation (Frescas and Pagano, 2008). Immunoprecipitation with cdk2 antibodies, followed by the addition of either recombinant Rb or histone H1 substrates and γ-ATP in in vitro kinase assays, confirmed the loss of cdk2 catalytic activity following homocysteine and K2 inhibitor II treatment. Thus, inhibition of cdk2 or cdk4 activity blocked cell cycle progression and correlated with increased cell survival in the presence of homocysteine treatment.
As an alternative to inhibit the G1 cdks, we attempted to reduce cyclin D1 expression by using antisense oligonucleotides (Fig. 5B). Sense and antisense oligonucleotides against cyclin D1 were transfected into differentiated neurons. Two days later cells were stained with anti-cyclin D1 antibodies and analysed by confocal immunofluorescence or harvested for immunoblot analysis with cyclin D1 antibodies (Fig. 5B, left). A significant reduction in cyclin D1 expression was detected by both methods and quantitated (Fig. 5B, left). After treatment with 0.25 mM homocysteine for 3 days, differentiated neurons transfected with antisense oligonucleotides showed significantly less apoptosis than cells that had been transfected with sense oligonucleotides (Fig. 5B, right). This was consistent with the idea that cyclin D1–cdk4 played an essential role in causing cell cycle re-entry and the concomitant neuronal apoptosis.
Other groups have suggested that the tyrosine kinase c-Abl becomes activated during the neuronal cell death induced by stressors such as β-amyloid fibrils (Alvarez et al., 2004; Cancino et al., 2008). To determine if Abl was involved in the regulation of homocysteine-dependent neuronal apoptosis, we examined untreated and homocysteine-treated cells by immunoblot analysis using Abl antibodies, and found that the expression of c-Abl increased with treatment time (Fig. 5C). We next treated neurons with Gleevec™ (STI571 mesylate salt), an inhibitor of c-Abl kinase activity (Avramis et al., 2003; Hagerkvist et al., 2007) before homocysteine treatment and measured cell survival. Treatment with Gleevec™ rescued neurons from homocysteine-induced cell death in a dose-dependent manner (Fig. 5D). Using immunofluorescent staining with multiple antibodies, we further evaluated the effects of Gleevec™ on G1 cell cycle activation (Fig. 5E). Gleevec™ treatment blocked the translocation of cyclin D1 and cdk4 from the cytoplasm to the nucleus in homocysteine-treated neurons (Fig. 5E), suggesting that c-Abl inhibition affected G1 phase progression. Gleevec™ treatment also reduced the detection of phospho-Rb after 7 h of homocysteine treatment (Fig. 5F), suggesting that these neurons had not progressed to nor activated the G1–S phase checkpoint. Thus, by chemical cdk inhibition of cdk2 and cdk4, decreasing cyclin D1 expression, or preventing cyclin D–cdk4 nuclear translocation via inhibition of c-Abl, we were able to block neuronal cell death, directly linking G1 cdk activation to homocysteine-induced apoptosis.
We wanted to determine whether blocking S phase progression was also sufficient to block homocysteine-induced death. We pretreated neurons with DNA replication inhibitors, such as aphidicolin, 1 mM hydroxyurea or 1 µM nocodazole, and 2 h later 0.25 mM homocysteine was added to the cultures as indicated (Fig. 6). Aphidicolin specifically inhibits DNA polymerase-α (Oguro et al., 1979; Kowalzick et al., 1982), hydroxyurea blocks ribonucleotide reductase (Adams and Lindsay, 1967) and nocodazole inhibits microtubule dynamics and arrests the cell cycle at G2/M phase. Unlike the G1 phase inhibitors, none of these DNA replication inhibitors prevented neuronal death induced by homocysteine treatment (Fig. 6). In fact, treatment of neurons with nocodazole was toxic and co-treatment with homocysteine increased cell death even further. This suggests that once a cell entered S phase and began to replicate DNA, apoptosis was not preventable.
The DNA damage response pathway is activated by both physical DNA damage (DNA lesions and/or double- or single-stranded DNA breaks) or in response to DNA replication stress, due to stalled replication forks. The DNA damage response has been shown to be essential for the neuronal apoptosis due to DNA damaging agents, such as camptothecin (Rich et al., 2000; Roos and Kaina, 2006). We wanted to determine whether homocysteine treatment also caused DNA damage in cortical neurons. Using the Comet assay (Angelis et al., 1999; Lemay and Wood, 1999), which detects DNA single-strand breaks and incomplete excision repair sites, we showed that damage was detectable within 4 h of homocysteine treatment (Fig. 7A). More than 70% of the homocysteine-treated cells stained positively in this assay, whereas only a few Comet positive cells (<8%) were detected in non-treated neurons (Fig. 7A). Phosphorylation of H2AX (γ-H2AX) is an early sign of DNA damage induced by stalling replication forks (Bartek and Lukas, 2007; Tanaka et al., 2007; Chanoux et al., 2009). By immunoblot analysis with γ-H2AX antibodies, we detected an increase in γ-H2AX following homocysteine treatment (Fig. 7B) suggesting that homocysteine-dependent DNA damage may be due to inappropriate DNA synthesis.
The central proteins of the DNA damage response are ATM and ATR (Rotman and Shiloh, 1999; Pandita, 2002; Brown and Baltimore, 2003). While ATM and ATR share most downstream effectors, the ATM-Chk2 pathway is primarily activated by DNA double-strand breaks and the ATR-Chk1 pathway responds to stalled DNA replication forks (Shiloh, 2003; Bartek and Lukas, 2007). To determine the role of the DNA damage response pathway in homocysteine-induced apoptosis, we treated neurons with homocysteine and Ku-55933, an inhibitor of ATM protein (Hickson et al., 2004) (Fig. 8A). We expected that if activation of the DNA damage response was responsible for triggering the cell death program, inhibition of ATM activity would block homocysteine-induced apoptosis. We found that Ku-55933 treatment did not rescue homocysteine-induced apoptosis, and in fact exacerbated it in a dose-dependent manner (Fig. 8A and B). Ku-55933 treatment in the absence of homocysteine had no affect on cell survival (Fig. 8A). Similar results were seen with caffeine treatment, which is a powerful inhibitor of ATR kinase (Sarkaria et al., 1999; Lu et al., 2008) (Fig. 8C).
By immunoblot analysis with anti-cleaved caspase-3 antibodies using lysates from treated and untreated neurons, we saw that homocysteine-induced caspase-3 activation 8 h post treatment (Fig. 8E). Co-treatment with homocysteine and K2 inhibitor II prevented the detection of activated caspase-3, consistent with the reduced apoptosis seen in the presence of this inhibitor (Fig. 5A). Co-treatment with homocysteine and Ku-55933 or homocysteine and caffeine increased caspase-3 activation (Fig. 8E), consistent with the exacerbation of homocysteine-dependent apoptosis caused by inhibition of the ATM/ATR pathway seen in these cells (Fig. 8A–C).
Thus, inhibitors of ATR/ATM did not rescue homocysteine-induced neuronal cell death, suggesting that the DNA damage response pathway was not inducing apoptosis and might, in fact, normally function to prevent or slow homocysteine-induced apoptosis. This was in contrast to what was seen in camptothecin treated neurons (Fig. 8D). Camptothecin is a topoisomerase I inhibitor that induces apoptosis of cultured cortical neurons in a DNA damage dependent manner (Morris and Geller, 1996; Staker et al., 2002; Zhang et al., 2006), and in our experiments, 10 µM camptothecin killed more than 40% of the treated neurons within 24 h (Fig. 8D). However, Ku-55933 treatment significantly enhanced cell survival levels, demonstrating that camptothecin-induced death was dependent on ATM activity and camptothecin-dependent activation of the DNA damage response pathway triggered apoptosis (Fig. 8D).
Activation of the ATR pathway was seen by immunoblot analysis using phospho-ATR antibodies that recognize the activated form of the protein, in both homocysteine and camptothecin-treated cells (Fig. 9A, lane 1; 9C, lane 1). ATM and ATR activation causes the phosphorylation and activation of Chk1 and Chk2 (Zhou and Elledge, 2000; Brown and Baltimore, 2003; Roos and Kaina, 2006; Pabla et al., 2008). Chk1 and Chk2 in turn can inhibit cell cycle progression, induce the expression of DNA repair enzymes, or ultimately mediate apoptosis if the damage is irreparable (Takai et al., 2000; Maude and Enders, 2005; Vitale et al., 2008). Immunoblot analysis with phospho-Chk1 or phospho-Chk2 antibodies was performed with lysates from neurons treated with homocysteine. Both Chk1 and Chk2 were phosphorylated in response to homocysteine treatment (Fig. 9A, lanes 2 and 3, and 9B). Interestingly, we found that the two checkpoint kinases have different kinetic responses during the time course of homocysteine treatment. Detection of phospho-Chk2 increased rapidly between 1 and 4 h of exposure (Fig. 9A, lane 3), but decreased by 8 h. Phospho-Chk1 was not increased until 8 h and continued to increase up to 24 h of treatment (Fig. 9A, lane 2).
Treatment with both homocysteine and caffeine, which blocks ATR activation, reduced the detection of phospho-ATR (Fig. 9C, lane 1), while Ku-55933 treatment, which primarily inactivates ATM, did not affect ATR activation. Treatment of neurons with homocysteine and Ku-55933, however, significantly reduced but did not eliminate, Chk1 phosphorylation (Fig. 9B), verifying that Ku-55933 prevented the activation of the homocysteine-dependent DNA damage response pathway activation. Phospho-ATR, phospho-Chk1 and phospho-Chk2 were also detected following camptothecin treatment (Fig. 9A, lanes 1–3). When neurons were treated with both Ku-55933 and camptothecin, phospho-ATR and phospho-Chk1 detection was reduced, but phospho-Chk2 activity was blocked completely (Fig. 9A, lanes 1–3).
The tumour suppressor p53 has been shown to play an essential role in the apoptosis induction seen in response to DNA damage in camptothecin treated cortical neurons (Martin et al., 2008). In the absence of DNA damage, p53 is poorly phosphorylated and rapidly degraded. In the presence of DNA damage, Chk1 and Chk2, among other proteins, can phosphorylate and stabilize p53. To determine whether p53 became phosphorylated and was stabilized in response to homocysteine-induced DNA damage, we used phospho-p53 antibodies in immunoblot analysis from lysates of treated neurons. Phosphorylation of p53 was not detected in untreated, differentiated neurons, but became weakly visible between 8 and 24 h of homocysteine treatment (Fig. 9A, lane 5). However, 8 h of campothecin treatment permitted the detection of strong p53 phorphorylation, which was lost in the presence of Ku-55933 treatment (Fig. 9A, lane 5). Thus, while DNA damage is seen in response to homocysteine treatment (Figs 7 and 9C, lane 5), apoptosis induction by homocysteine seems p53-independent, and is distinct from the damage induced by campothecin.
Homocysteine treatment appeared to activate the DNA damage response pathway as seen by phosphorylation of ATR, Chk1 and Chk2, and Ku-55933 or caffeine treatment appeared to reduce this activation. Blocking the DNA damage response pathway by either Ku-55933 or caffeine, however, was unable to rescue neurons from homocysteine-induced death (Fig. 8) and in fact exacerbated neuronal apoptosis. We wanted to determine whether blocking the DNA damage response pathway had any affect on cell cycle reactivation. Lysates from neurons treated with homocysteine and different inhibitors (Ku-55933, caffeine or K2 inhibitor II) were examined by immunoblot analysis to examine the levels of the G1 cell cycle proteins. Ku-55933 or caffeine pre-treatment did not block homocysteine-induced G1 cell cycle progression or activation: p27 levels were reduced (Fig. 9C, lane 2) and Rb became phosphorylated (Fig. 9C, lane 4), suggesting that homocysteine treatment initiated cell cycle re-entry independently of the activation status of the DNA damage response pathway. This was in contrast to what was seen when neurons were treated with homocysteine and K2 inhibitor II. As shown in Figs 4B and C and 5A, and again in Fig. 9C, pre-treatment with K2 inhibitor II prevented homocysteine-induced cell cycle reactivation and blocked neuronal cell death. The homocysteine-dependent reduction in p27 levels was not detected (Fig. 9C, lane 2), and phospho-Rb was only weakly detected compared to the levels seen with homocysteine alone, homocysteine and Ku-55933 or homocysteine and caffeine (Fig. 9C, lane 4). Without cell cycle re-entry (no homocysteine or homocysteine and K2 inhibitor II treatment), H2AX phosphorylation was reduced, compared to levels seen in homocysteine alone or with homocysteine and Ku-55933 (Fig. 9C, lane 5). H2AX phosphorylation was detected even in the presence of Ku-55933 treatment (Fig. 9C, lane 5), as it can be phosphorylated by other replication stress-induced kinases, such as ATR or DNA-dependent protein kinase (Park et al., 2003; Chanoux et al., 2009). The homocysteine and K2 inhibitor II treated neurons only partially activated the DNA damage response pathway. Phospho-ATR was detected by immunoblot analysis (Fig. 9C, lane 1), but Chk1 phosphorylation was prevented by the presence of the K2 inhibitor II (Fig. 9B). While ATR was activated in response to homocysteine treatment, in the absence of cdk2 activity and cell cycle progression, it appeared unable to propagate the complete DNA damage response signal.
The reappearance of cell cycle markers is one of the earliest abnormalities seen in affected neurons in patients with Alzheimer’s disease, but the significance of this remains unclear (Webber et al., 2005; Herrup and Yang, 2007). Investigation into the pathways that stimulate cell cycle re-entry is both necessary to elucidate this process as well to provide potential therapeutic targets. We demonstrated that homocysteine treatment of differentiated, quiescent cortical neurons in culture caused reactivation of the G1 cell cycle proteins, permitting transit into S phase. Homocysteine treatment caused cyclin D1, cdk4 and cdk2 nuclear translocation and down-regulation of the cdk inhibitor p27Kip1. Reactivation of the cell cycle machinery was required for homocysteine-induced apoptosis as blocking the G1 cell cycle proteins, by a variety of methods, prevented neuronal cell death. Cell cycle re-entry permitted entry into S phase and subsequent DNA synthesis, and it appeared that S phase entry was required to trigger apoptosis. We observed a 95% co-localization of DNA synthesis and apoptosis markers in homocysteine-treated neurons, suggesting that cells that started DNA replication ultimately died. However, blocking cell cycle progression post G1, using DNA synthesis inhibitors, was unable to prevent apoptosis, suggesting that death was not preventable after the G1–S phase checkpoint.
Interestingly, the G0–G1–S phase transition seen in cultured neurons remarkably resembles the mitogen-dependent G0–G1–S phase transition seen in epithelial or lymphoid cells (Sherr and Roberts, 1999; Becker and Bonni, 2004; Bockstaele et al., 2006). In both scenarios, cyclin D1 moves to the nucleus, where it partners with cdk4. Cdk4 and cdk2 phosphorylate Rb, which permits entry into S phase and DNA replication initiation. The outcome, however, is different and suggests that differentiated neurons may no longer be competent to complete S phase or the entire cell cycle. Apoptosis is required during the development of the nervous system (Becker and Bonni, 2004, 2005). However, DNA synthesis and activation of cell cycle proteins is not detected in apoptosis in the developing brain, suggesting that DNA synthesis in ‘post-mitotic’ neurons may be a death signal.
How homocysteine reactivates the cell cycle is not clear, but direct effects on cyclin D1 and p27 seem likely. These are critical sensors in epithelial and lymphoid cells, sensitive to a variety of soluble mitogenic and antimitogenic factors, as well as cell–cell and cell–extracellular matrix interactions (Nickeleit et al., 2007; Malumbres and Barbacid, 2009). Proteins, such as glycogen synthase kinase 3-β (Pontano and Diehl, 2009) or S-phase kinase-associated protein 2 (Frescas and Pagano, 2008), which regulate mitogen-dependent cyclin D1 nuclear translocation and p27 degradation, respectively, in these cell types, might be involved in homocysteine-dependent cell cycle re-entry as well. We were able to block homocysteine-dependent apoptosis by inhibition of G1 cell cycle progression at multiple different steps, stressing the importance of the entire G1 cell cycle phase rather than the significance of any one protein, and highlighting the plethora of potential therapeutic targets this pathway presents. This is consistent with the proposed linearity and temporal order of the ‘Rb pathway’ during the G0–G1 phase transition in epithelial or lymphoid cells, where cyclin D–cdk4 may be required to activate cdk2, which in turn may be required to target p27 for degradation; and the combined contribution of both cdk4 and cdk2 may be needed for full Rb phosphorylation and S phase entry (Blain, 2008; Malumbres and Barbacid, 2009).
In our model we tested a physiological Alzheimer’s disease risk factor, homocysteine, which is produced as a natural by-product of methionine metabolism. Increased levels of total plasma homocysteine can be caused by genetic defects, nutritional deficiencies in folate, vitamin B12, vitamin B6 or other factors (Welch and Loscalzo, 1998; Gueant et al., 2005). Elevated plasma homocysteine concentrations were found in patients with chronic renal failure (Wilcken and Gupta, 1979; Chauveau et al., 1993), hypothyroidism (McCully, 1996) and pernicious anaemia (Savage et al., 1994). Hyperhomocysteinaemia was shown to be a strong, independent risk factor for dementia and Alzheimer’s disease (Seshadri et al., 2002; Mizrahi et al., 2004; Agnati et al., 2005; Ravaglia et al., 2005; Blasko et al., 2008), Parkinson’s disease (Mattson and Shea, 2003) and stroke (Hankey and Eikelboom, 2001). Elevations in plasma homocysteine levels precede the development of dementia and there is a continuous, inverse linear relation between plasma homocysteine concentrations and cognitive performance in aged adults (McCaddon et al., 1998; Irizarry et al., 2005; Kado et al., 2005). Hyperhomocysteinaemia also appears to exacerbate the effects of other factors implicated in Alzheimer’s disease pathology, such as increasing tau phosphorylation and beta-amyloid toxicity (Pacheco-Quinto et al., 2006; Luo et al., 2007). Neuronal apoptosis has been detected in the hippocampal and cortical neurons of folate deficient mouse models (Kruman et al., 2000, 2005). While others have studied the effects of genotoxic or DNA damaging agents as inducers of neuronal apoptosis, we chose homocysteine not only for its physiological significance but because its ability to inflict DNA damage was more subtle. Extended hyperhomocysteinaemia in patients may cause a similar cell cycle re-entry in affected neurons, which might eventually contribute to neuronal cell loss.
DNA damage occurs in response to toxic chemicals, as a byproduct of normal metabolism through the generation of reactive oxygen species or free radicals, or through errors in DNA replication, such as DNA replication fork stalling due to low nucleotide pools or blocking DNA lesions. Our results demonstrated that homocysteine did cause DNA damage as seen by increases in Comet positive neurons. However, the appearance of γ-H2AX suggests that homocysteine-induced DNA damage may be due to replication stress rather than more traditional DNA breaks, such as caused by DNA damaging agents like camptothecin. Homocysteine activates the DNA damage response, as seen by phosphorylation of ATR, Chk1 and Chk2, but the DNA damage response induced by homocysteine appears different than the DNA damage response induced by camptothecin. The homocysteine-induced DNA damage response appears to be p53-independent and inhibition of ATM/ATR activity does not block apoptosis, but rather exacerbates it. This is in stark contrast to the actions of camptothecin, which appears to induce p53, and be ATM/ATR dependent: apoptosis is blocked when these proteins are inhibited. The 95% colocalization of S phase markers and apoptotic markers suggests that homocysteine-induced apoptosis is due to S phase entry, and subsequent DNA replication stalling might be the cause of DNA damage, activating the DNA damage response. The activation of the DNA damage response pathway was not responsible for homocysteine-dependent neuronal cell death, as inhibition of this pathway accelerated the induction of apoptosis, suggesting that it may have a more neuroprotective function. Support for this idea comes from recent studies that have shown that the ATR-Chk1 pathway can protect HCT116 cells from caspase-3-dependent apoptosis, suggesting that the pathways that protect from replication stress are different than those induced by more traditional DNA damaging agents (Myers et al., 2009). Others have demonstrated that agents like etoposide or methotrexate (Kruman et al., 2004), camptothecin (Martin et al., 2008; Tian et al., 2009), ultraviolet irradiation (Park et al., 1998), γ-irradiation (Enokido et al., 1996) and arabinoside (Tomkins et al., 1994) directly and irreparably damage DNA and induce apoptosis. Our data suggest that homocysteine causes apoptosis by a different mechanism, directly causing the G0–S phase transition, with the DNA damage response pathway only activated to block the S phase-dependent apoptotic response.
While others have suggested that DNA damage itself is the causative agent of cell cycle reactivation and DNA synthesis (Kruman et al., 2004; Tian et al., 2009), in our system, we found that blocking the DNA damage response pathway with Ku or caffeine treatment, did not block cell cycle re-entry, as seen by p27 down-regulation and Rb phosphorylation. Neurons still entered S phase, but in the absence of the protective effect of the DNA damage response, apoptosis was increased. Conversely, blocking cell cycle re-entry with K2 inhibitor II also decreased γ-H2AX detection, presumably due to decreased DNA synthesis, but did not block ATR activation. Clearly, the two pathways are interconnected: DNA synthesis increases DNA damage that activates the DNA damage response pathway that feedbacks and attempts to halt and/or repair synthesis-induced damage, but at least to some extent they function independently. We would suggest that the nature of the initiating agent, homocysteine versus camptothecin, and the type of ensuing DNA damage (replication stress versus double-strand breaks) might explain the difference between our results and those of others.
The parallels between mitogen-induced G0 exit in epithelial and lymphoid cells and differentiated neurons should not be overlooked. We were able to block homocysteine-induced death only by blocking G1 phase progression. Once the neuron had reached S phase and began to replicate DNA, apoptosis could not be prevented and activation of the DNA damage response may only be prolonging an inevitable death. The appearance of cell cycle markers in neurodegenerative disease in vivo might represent a state that is already retractable to intervention, and the real challenge may be to prevent the initial G1–S phase progression. Cell cycle proteins have been therapeutic targets in cancer drug development for decades and it will be interesting to determine whether these types of inhibitors might be useful in the treatment of neurodegenerative disease as well.
American Cancer Society (to S.W.B.); and a National Institute of Aging Leadership Award in Alzheimer’s Disease Research (grant number 5 K07 AG00959 to Dr Suzanne Mirra).
Gleevec™ (STI571 mesylate salt) was a kind gift from Novartis Pharmaceuticals. We thank Dr Ting-Chung Suen and Dr Sarah Nataraj for helpful discussion, and Dr William Oxberry for help with confocal microscopy techniques.
ataxia-telangiectasia mutant protein
ataxia telangiectasia and Rad3 related protein
checkpoint protein kinase
terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labelling