Bmi1 Is Down-Regulated in the Aging Brain and Displays Antioxidant and Protective Activities in Neurons

Aging increases the risk to develop several neurodegenerative diseases, although the underlying mechanisms are poorly understood. Inactivation of the Polycomb group gene Bmi1 in mice results in growth retardation, cerebellar degeneration, and development of a premature aging-like phenotype. This progeroid phenotype is characterized by formation of lens cataracts, apoptosis of cortical neurons, and increase of reactive oxygen species (ROS) concentrations, owing to p53-mediated repression of antioxidant response (AOR) genes. Herein we report that Bmi1 expression progressively declines in the neurons of aging mouse and human brains. In old brains, p53 accumulates at the promoter of AOR genes, correlating with a repressed chromatin state, down-regulation of AOR genes, and increased oxidative damages to lipids and DNA. Comparative gene expression analysis further revealed that aging brains display an up-regulation of the senescence-associated genes IL-6, p19Arf and p16Ink4a, along with the pro-apoptotic gene Noxa, as seen in Bmi1-null mice. Increasing Bmi1 expression in cortical neurons conferred robust protection against DNA damage-induced cell death or mitochondrial poisoning, and resulted in suppression of ROS through activation of AOR genes. These observations unveil that Bmi1 genetic deficiency recapitulates aspects of physiological brain aging and that Bmi1 over-expression is a potential therapeutic modality against neurodegeneration.


Introduction
Aging is the prime risk factor for the development of several neurodegenerative diseases. Although the reason for this is unknown, it is possible that specific neuronal genes having protective roles in the aging brain are down-regulated due to increased mutational rate and/or defective DNA repair at their promoter [1]. Moreover, long lived neuronal cells are more likely to accumulate mutations in their genomic DNA than most other cell types with age, leading to impaired cellular functions [2,3]. This phenomenon may be explain in part due to the inability of post-mitotic cells to replicate their DNA, a process that is tightly coupled to DNA damage checkpoint and DNA repair [4,5,6]. Moreover, neurons possess a high metabolic activity and consume large amount of O 2 , and thus are exposed to higher levels of oxidative stress compared to other tissues [3,6].
Mitochondria are not only cellular organelles required for aerobic respiration but are also the main source of intracellular reactive oxygen species (ROS), which are thought to be causal for most oxidative damage accumulation with age [3,7,8]. The balance between ROS and antioxidant molecules is critical to determine the rate of oxidative damage accumulation, and thus possibly cellular and organism lifespan [9]. The free radical theory of aging is partially supported by gain-and loss-of-function experiments showing that antioxidant-encoding genes such as catalase, Cu/Zn superoxide dismutase (Sod1) and Mn superoxide dismutase (Sod2) are critical determinants of lifespan [10,11,12,13]. Thus, reduced antioxidant response (AOR) genes expression in the aging brain could displace the oxidative balance and may instigate accelerated aging and age-related neurodegenerative diseases.
Polycomb group proteins form large multimeric complexes that silence specific target genes by modifying chromatin organization [14]. The Polycomb group protein Bmi1 is a component of the Polycomb Repressive Complex 1 (PRC1), which promotes chromatin compaction and gene repression through its histone H2A E3-mono-ubiquitin ligase activity [15,16,17]. Bmi1 2/2 mice present axial skeleton defects, reduced post-natal growth and lifespan, along with progressive cerebellar degeneration [18,19]. Yet, we previously reported that Bmi1 2/2 mice develop a progeroid phenotype in the central nervous system characterized by lens cataracts, cortical neurons apoptosis, p53 activation and oxidative damage accumulation [20]. Bmi1 2/2 neurons are also hypersensitive to mitochondrial toxins, to DNA-damaging agents and to intracellular b-amyloid peptides, and display elevated ROS concentrations due to severely reduced AOR genes expression. Notably, most of Bmi1 functions on the oxidative metabolism of neurons can be explained by p53-mediated repression of AOR genes expression [20,21].
In this study, we investigated the possible molecular links between physiological brain aging and Bmi1 genetic deficiency, and analyzed Bmi1 neuroprotective potential. We show that in the aging brain, where Bmi1 expression declines in the neurons, there is an up-regulation of inflammatory and senescence markers, accumulation of p53 to AOR gene promoters, reduced expression of AOR genes and elevated oxidative damage to neurons, in a pattern that is similar to that of Bmi1 2/2 mice. Increasing Bmi1 expression in cortical neurons was highly neuroprotective against inhibition of the mitochondria respiratory chain or the DNA topoisomerase I, and resulted in activation of antioxidant defenses and in suppression of ROS levels. This work suggests that modulation of Bmi1 expression in the aging or disease-affected brain could provide robust neuroprotective activity.

Ethical statement
Animals were maintained in our facilities and used in accordance with and after approval by the Animal Care Committee of the Maisonneuve-Rosemont Hospital Research Center (Approval ID #2008-27 and #2009-42). Paraffin-embedded archival brains were obtained from the department of pathology of the Maisonneuve-Rosemont Hospital. Post-mortem human eyes were provided by the Eye's Bank of our institution and used with permission of Maisonneuve-Rosemont Hospital ethical committee.

Animals
Bmi1 +/2 mice (in C57Bl/6J background) were a gift from M. Van Lohuizen (The Netherlands Cancer Institute, Amsterdam). Wild type C57Bl/6J inbred mice were purchased from Charles River Canada.
Immunohistochemical analysis, quantitative real-time PCR, Western blot, measurement of ROS and lipid peroxidation levels Technical details are provided in Information S1 [22].

Statistical analysis
Statistical differences were analyzed using Student's t-test for unpaired samples. An analysis of variance (ANOVA) followed by the Dunnett test was used for multiple comparisons with one control group. In all cases, the criterion for significance (P value) was set as mentioned in the figures.

Bmi1 is down-regulated in mouse cortical neurons during aging
To evaluate Bmi1 expression in neurons during aging, we analyzed sections of cerebral cortices from young (2 months) and old (21-26 months) mice using immunohistochemistry. We observed that Bmi1 expression decreased significantly in old NeuN + cortical neurons, but not in GFAP + astrocytes ( Figure 1A). In old neurons, Bmi1 labeling was not uniform, with some neurons expressing Bmi1 at moderate levels, while others showing nearly undetectable levels ( Figure 1A). Immunolabeling was not present in cortical sections from Bmi1 2/2 mice, confirming the specificity of the antibody ( Figure S1). Referring to a previous report proposing that Bmi1 is not down-regulated in the aging mouse brain [23], we measured Bmi1 levels in a large cohort of young (n = 15) and old (n = 15) inbred mice. Using whole cerebral extracts and quantitative RT-PCR (Q-PCR), we found that Bmi1 mRNA levels decrease by ,60% in old cortices, thus suggesting reduced Bmi1 transcription ( Figure 1B). Interestingly, Bmi1 expression levels were highly variable in the old age population, with some mice showing only 20% of Bmi1 levels compared to young mice ( Figure 1B). To confirm our observations at the protein levels, we perform Western blot analysis of whole cortices from young and old mice, which revealed a ,40% reduction in Bmi1 levels in old mice ( Figure 1C). To address the possibility that a variation in cell population numbers could underlies the observed reduction in Bmi1 levels, we counted and compared the number of NeuN + neurons/cortical area or over the total number of cortical cell nuclei in young and old mice, but found no difference ( Figure 1D).
p16 Ink4a and p19 Arf are the main targets of Bmi1 transcriptional repressive activity in most of cell types, including cortical neurons [18,20]. Expression of p16 Ink4a and p19 Arf was increased concomitantly with Bmi1 reduction in old brains ( Figure 2A). The similarity between normal brain aging and Bmi1 genetic deficiency also extended to Noxa, a p53-regulated gene involved in apoptosis, and IL-6, which is involved in inflammation and inflammatory senescence ( Figure 2A) [24,25]. Infection of cultured Bmi1 2/2 neurons with a dominant negative adenovirus against p53 (DNp53) could restore IL-6 expression to normal levels, but not those of p16 Ink4a and p19 Arf ( Figure S2), which are primary Bmi1 targets [18,20,26]. To further seek for possible gene expression pattern resemblances between brain aging and Bmi1 genetic deficiency, we compared the expression of AOR genes in cortices of both young and old mice, and of post-natal day (P) 25 WT and Bmi1 2/2 mice. Notably, several genes involved in ROS scavenging were down-regulated in old brains, revealing a general decrease in antioxidant defenses ( Figure 2B). Common downregulated genes between Bmi1 2/2 brains and old brains were xCT, NQO1 and Sod2 ( Figure 2B). These genes are also downregulated in cultured Bmi1 2/2 neurons [20]. To commensurate the above observations, we compared the accumulation of oxidative damages between young and old brains, and between WT and Bmi1 2/2 P25 brains or cultured neurons. Concentrations of oxidized lipids were increased in old brains (1.4060.05, P,0.05, n = 3) compared to the young samples, similarly as between Bmi1 2/2 (1.4360.03, P,0.05, n = 3) and WT brains or Bmi1 2/2 (2.5460.12, P,0.01, n = 3) and WT e18.5 neurons. Concomitant with this finding, accumulation of oxidized genomic DNA bases (8-hydroxydeoxyguanosine) was elevated in neurons of old and P25 Bmi1 2/2 brains ( Figure 2C) [27,28]. In cortical neurons, 8-oxoguanine lesions were found in both nuclear and mitochondrial DNA ( Figure 2C). This latter finding implies increased oxidative DNA damages in cortical neurons from old and P25 Bmi1 2/2 mice owing to reduced antioxidant defenses. This primary defect may be combined with a reduction in 8oxoguanine DNA glycosylase (OGG1) activity or level [29,30,31]. It was previously showed that p53 proteins levels progressively increase in the aging rat brain [28]. Furthermore, p53 protein can associate with the promoters of AOR genes such as those encoding glutathione S-transferase alpha 1 (GSTa1), NAD(P)H quinone oxidoreductase (NQO1) and xCT. Upon damage, activated p53 can repress the transcription of these genes and interfere with the antioxidant defense system normally activated by the Nrf2 transcription factor [32]. To test for a possible mechanism of AOR genes downregulation in the aging brain, we compared young and aged brains for p53 accumulation at the xCT and Sod2 promoters by ChIP. We found that p53 was enriched at the xCT promoter and in two distinct Sod2 promoter regions in aged brains ( Figure 2D). Interestingly, this correlated with a repressed chromatin state, as shown by the accumulation of heterochromatin marks at these promoters, similarly as found in cultured Bmi1 2/2 neurons ( Figure 2D) [20]. To test if age-dependent Bmi1 down-regulation was restricted to mice, we analyzed BMI1 expression in young and old human brains by immunohistochemistry. We observed that BMI1 expression was reduced in hippocampal neurons of old brains ( Figure 3A). In contrast to the observations found in mice, BMI1 is not expressed in normal human astrocytes, even in young brains [20]. We previously reported on Bmi1 expression in post-mitotic neurons of the mouse retina. Using immunofluorescence, we analyzed BMI1 expression in the adult human retina. We observed robust BMI1 immunoreactivity in the nucleus of photoreceptors and weaker expression in neurons of the inner nuclear and ganglion cell layers ( Figure 3B). The peculiar expression of BMI1 in some photoreceptors was distinct from classical polycomb bodies staining ( Figure 3B). To quantify for possible BMI1 down-regulation in the aging human retina, we performed Western blot analysis using young (4-30 years old), middle-aged (40 and 50 years old) and old (65-75 years old) human retinas. BMI1 protein levels were variable from sample to sample, especially in the young and middle-aged groups, in agreement with previous studies [1]. Nonetheless, BMI1 levels were consistently lower in old human retinas ( Figure 3C). Linear regression analysis suggests a stepwise reduction in BMI1 levels, and where expression between 85-95 years of age could be, on average, less than 30% of what is found in young individuals. However, due to samples limitations, BMI1 expression in retinas older than 75 years could not be tested in this study. Reactive gliosis in the retina and in the brain is observed in neurodegen-erative diseases, CNS lesions and during aging and can be visualize by GFAP immunoreactivity. P16 INK4A represents a biomarker of tissue aging and is a target of Bmi1 repressive activity [23,33,34]. Consistently, we observed increased GFAP and P16 INK4A immunoreactivity in sections of aged human retinas ( Figure 3D), similarly as observed in the retina and the brain of Bmi1 2/2 mice at P25 [20,35].

Bmi1 over-expression is neuroprotective and activates antioxidant defenses
To test if increasing Bmi1 expression could protect neurons from damage-induced cell death, we generated a plasmid expressing the complete human BMI1 cDNA ( Figure 4A). This expression vector was electroporated in e18.5 WT cortical neurons that were thereafter exposed or not to neuronal toxins ( Figure 4B). Using immunohistochemistry, neurons viability was determined by the absence of activated caspase-3 and the maintenance of long axons and dendrites presenting a rich arborization, similarly as observed in vehicle-treated neurons. When neurons were exposed to 5 mM of camptothecin (CA), a DNA topoisomerase I inhibitor that leads to single-strand DNA breaks formation and p53mediated neuronal apoptosis, only ,40% of control-GFP electroporated neurons were viable after 16 hours [24,36,37,38,39]. In contrast, ,80% of GFP/BMI1-transfected neurons remained viable after this time point (Figure 4C and D). These data revealed that BMI1 is required and sufficient to block CA-induced apoptosis cell autonomously. To test if BMI1 overexpression was neuroprotective against a mitochondrial toxin, we treated neurons with 8 mM of 3-nitroproprionic acid (3-NP), a succinate dehydrogenase inhibitor that blocks complex II of the mitochondria respiratory chain [40,41]. 3-NP treatment induced caspase-3-dependent (apoptosis) and caspase-3-independent (vacuolization) neuronal cell death ( Figure 4C). At this high concentration, only ,21% of GFP-electroporated neurons remained viable after 16 hours. In contrast, ,77% of GFP/ BMI1-transfected neurons remained viable at this concentration, showing that BMI1 over-expression confers cell autonomous protection against 3-NP toxicity ( Figure 4C and D). We previously reported that p53 protein levels were increased in Bmi1 2/2 neurons and that the major effects of Bmi1 genetic deficiency were mediated by p53 pro-apoptotic and pro-oxidant activity. To test part of the mechanism involved in BMI1 neuroprotective activity, we over-expressed BMI1 by DNA electroporation in WT and p53 2/2 neurons treated or not with 5 mM of CA. We observed that although p53 2/2 neurons were highly resistant to CA treatment compared to WT neurons, this effect was further enhanced when BMI1 was over-expressed ( Figure 4E).
To address if BMI1 activity correlated with ROS modulation and induction of AOR genes expression, a BMI1-expressing adenovirus was used to infect cortical neurons in culture. 48 hours post-infection, neurons were treated or not with 8 mM of 3-NP and analyzed 6 hours later for ROS concentrations. We found that BMI1 over-expression reduced ROS concentrations in untreated neurons ( Figure 5A). This biochemical effect was reflected by an increase in phase II genes (i.e. xCT, NQO1 and GST-1a) expression by an order of 3-9 fold and induction of Sestrin1 and Sod2 by an order of 1.5-2 fold ( Figure 5B). Upon 3-NP exposure, we also observed a reduction in ROS concentrations in BMI1-overexpressing neurons relative to GFP-infected neurons, and this was again reflected by activation of antioxidant genes ( Figure 5A and B). We concluded that in gain-of-function experiments, BMI1 can reduce ROS levels and over-activate antioxidant defenses. Notably, p53 2/2 neurons infected with the BMI1-expressing adenovirus or control adenovirus (GFP-only) showed nearly identical ROS concentrations, and these were comparable to ROS concentrations found in WT neurons infected with the BMI1-expressing adenovirus ( Figure 5C). These later results suggest that parts of BMI1 neuroprotective activities may be p53-independent and not mediated through regulation of ROS concentrations.

Discussion
Herein, we demonstrated that the Polycomb group gene Bmi1 is down-regulated in the brain and the retina of aged mice and humans. Moreover, numerous aspects of brain aging, but not all, are recapitulated in the progeroid phenotype of Bmi1 2/2 mice, suggesting a model linking Bmi1 deficiency and unregulated p53 activity to reduced antioxidant defenses in the aging brain. On the other hand, we presented evidences that BMI1 over-expression is neuroprotective, can activate AOR genes, and can suppress ROS.
In the central nervous system of mammals, peroxisome proliferator-activated receptor gamma coactivator-1a (PGC-1a) has been shown to be a major inducer of the antioxidant defense system. Consistent with this, dopaminergic neurons from PGC-1anull mice are hypersensitive to the mitochondrial complex I inhibitor MPTP, and PGC-1a over-expression is neuroprotective against paraquat and H 2 O 2 [42]. Likewise, Nrf2 over-expression in astrocytes can protect neurons in a non-cell autonomous manner from 3-NP toxicity, while Sirt1 over-expression can delay axonal degeneration and counteract neurodegeneration in a mouse model of Alzheimer disease [40,41,43,44]. However, with the exception of a modest reduction in Nrf1 mRNA levels, the expression of PGC-1a, Sirt1 and Nrf2 was unaltered at the mRNA and protein levels in Bmi1 2/2 mice [20]. Hence, the neuroprotective effects of BMI1 over-expression against CA and 3-NP toxicity are unlikely to be mediated through modulation of these factors. Rather, we propose that most of these effects are mediated through repression of p53 activity. In support of this 1) AOR genes expression profile of BMI1-overexpressing neurons is similar to that of p53-null neurons, and 2) p53 is the central mediator of CAinduced apoptosis in neurons [20,24,36,37,38,39]. On the other hand, we also showed that BMI1 over-expression in p53 2/2 neurons confers increased neuronal survival after CA treatment, but not reduction in ROS concentrations. Our finding suggests that BMI1 neuroprotective activities are not only mediated via p53-dependent and ROS-dependent pathways. Although numer-ous possibilities exist, this could be achieved through BMI1 repression of alternate tumor suppressors and/or improved DNA damage response and repair [45,46,47]. Notably, the mechanism by which Bmi1 regulates p53 activity remains elusive, although repression of the p19 Arf /Mdm2/p53 pathway through direct transcriptional repression of p19 Arf by Bmi1 remains a possibility.
The contexts of Bmi1 deficiency in mice (which leads to a severe and early lethal pathology) and of physiological aging are completely different. Hence, it is most unexpected that Bmi1 down-regulation in neurons during brain physiological aging leads to a molecular phenotype that resembles Bmi1 genetic deficiency. The apparent conservation of the Bmi1-p53 antioxidant relationship during physiological aging is thus an important finding that brings new knowledge on the mechanisms regulating the process of physiological aging in the central nervous system. Our previous findings provided a molecular mechanism to explain why Bmi1null neurons age prematurely and are hypersensitive to various stressors [20]. Considering Bmi1 down-regulation in the aging central nervous system, this may also explain why aged neurons are more sensitive to various insults. Cellular and DNA damage produced by mitochondrial ROS have been proposed to be the causal factor of cellular and organismal aging [1,3,7]. Herein we observed that p53 accumulates at AOR gene promoters in the aging mouse brain, correlating with reduced AOR genes expression. This suggests that deficient BMI1 expression in the human brain may contribute to reduce antioxidant defenses, leading to accumulation of oxidative damage and neuronal dysfunction. Although controversial, neuroinflammation is potentially a driving force of normal brain aging and neurodegenerative diseases. The cytokine IL-6 is a biomarker of brain aging, and high levels of IL-6 are found in senescent cells and Alzheimer disease brains [25,48]. We found that old and Bmi1 2/2 brains expressed high levels of IL-6, and that this was p53-dependent in cultured Bmi1 2/2 neurons. These observations suggest that Bmi1 deficiency during aging could amplify p53 activity and thus accelerate the process of normal brain aging and age-related neurological diseases in readily susceptible individuals (see proposed model in Figure 6). Finally, while increased p53 activity in aging neurons may be in part explained by the loss of Bmi1 expression, the ageassociated mechanism leading to Bmi1 down-regulation remains to be investigated. Although this mechanism apparently operates at the transcriptional level, as reported for other genes in the aging brain [1], age-dependent Bmi1 mRNA degradation through specific miRNA activity is possible [49]. Also, it cannot be excluded that activated p53 represses Bmi1 transcription, although such evidences do not exist yet. At last, Bmi1 protein may be also preferentially targeted for degradation in aging neurons [50,51,52]. In old mice, Bmi1 expression declined only in neurons, not in astrocytes. Notably, only astrocytes have the potential to generate tumors. These data indicate that down-regulation of the Bmi1 oncogene in neurons during aging is apparently not a general response to prevent cancer (at least in mice). Rather, it may represent an epigenetic change related to normal brain aging [1]. Taken together, our work suggests that targeted pharmaceutical stimulation of Bmi1 expression in aging neurons could prevent accumulation of age-associated oxidative damages, protect against apoptosis and possibly counteract the progression of neurodegenerative diseases. Figure S1 Coronal sections of WT and Bmi1 2/2 mice brains were analyzed by IHC using Bmi1 (brown) and